Responses to Geocoding ??
Chrissyvan Hilst
CVANHILS at SIVM.SI.EDU
Mon Jun 3 15:28:59 CDT 1996
EMAIL RESPONSES TO GEOCODING QUESTION
These are the reponses to the post I put on TAXACOM and FWIM, inquiring
about thoughts other museums or individuals had about dealing with
"value added" lat/long data attached to specimen localities. This is an
extensive post, I apologize, but the responses were informative and good
reading so I felt it necessary to share.
Thank you to everyone who responded,
Chrissy van Hilst
From: "Rod Tuloss" ret at pluto.njcc.com
There is a worldwide data base of military origin that will give a
lat/long in exchange for a place name. The list of place names (I
understand) is quite extensive including geologic/hydrologic
features as well as bridges, roads, railways, etc.
From: "Melissa C. Winans" <mcwinans at MAIL.UTEXAS.EDU>
I agree with you that the precision of your data is a serious
issue, and that it is important to do something to inform the user
of differences in precision from site to site. We had to deal with
the same question when we set up a GIS for our fossil specimen
sites about 4 years ago; some sites (especially the ones from the
late 1800's and early 1900's) could not be spotted precisely due to
inadequate data, and others covered a large enough area that it was
not possible to represent them with a single point at the 1:24,000
map resolution we were using. We solved this by doing three
things: 1) We marked sites that were too large to represent as a
single point with polygons outlining the extent of the site; 2) for
sites where the location was uncertain, but an educated guess could
be made, we outlined the most probable area and digitized that as
a polygon; 3) we attached an attribute to each point or polygon
indicating the degree of reliability of the plotting on a 5-step
scale ranging from "Exact" at the top to "Wild guess" at the
bottom.
This still leaves you with the question of what to do about
coordinates for the sites that can't be represented as points.
Depending on how you do your distribution analyses, you still may
have to represent the large or uncertain sites by a single
coordinate pair (probably the centroid of the polygon), but if your
GIS attribute tables have a reliability attribute you at least will
be able to take whatever steps seem to be called for to indicate
the degree of certainty of each location.
****************************************************************
Melissa C. Winans, Collection Manager (mcwinans at mail.utexas.edu)
Vertebrate Paleontology Laboratory Phone: 512-471-6087
J.J. Pickle Research Campus Fax: 512-471-5973
University of Texas, Austin, TX 78712
From: Don Kirkup <d.kirkup at LION.RBGKEW.ORG.UK>
Organization: The Royal Botanic Gardens, Kew, UK
We have c. 15,000 collections records in the Brunei Checklist
database (MS ACCESS). Most of the records are georeferenced and
each set of coordinates is assigned an error-code (in meters)
depending on how they were derived. Our scale of errors varies from
c. +/-100m for GPS derived coordinates, to locations given to the
nearest minute, to gazetteer derived localities for which we assign
an arbitrary error of +/-2,500m. We also have some collections for
which the only geographical information is 'Brunei', but thankfully
not too many! For a pictorial representation of species ranges a
map with different symbols for the different error codes is
probably enough to convey the message. For analytical purposes
however we felt that the georeference could not satisfactorily be
treated as a point, but should be treated as a polygon.
The first step of our approach was to simply draw a circle with a
radius for each collection point equivalent to that of the
error-code value in the database.
The precision of a locality was then further refined by finding the
intersection of the error-circle with graphical coverages related
to observations in our database. For example, collection
observations in the database includes forest and soil types. We
also have graphical coverages for both of these. The refined
locality is calculated by overlaying the error-circle with the
forest type and soils coverages and finding the intersection
between these three graphical layers and the tabulated values in
the database. For collections along 'linear' features such as roads
and rivers rather than draw an error-circle, we create a 50m buffer
around the feature then overlay and link to the database in the
same way.. This involves a fair amount of work even though we have
managed to semi-automate the process (using PC and Workstation
ArcInfo). We felt that this was the best we could achieve and this
was necessary if we were to incorporate as many collection
observations as possible for the interpretation of our
remote-sensed data.
Don Kirkup
Brunei Checklist Project
Herbarium
RBG Kew.
From: Jorge Soberon Mainero
<jsoberon at MIRANDA.ECOLOGIA.UNAM.MX>
The subject of geocoding label information is really huge. In
CONABIO we have been helping Mexican taxonomists to geocode tens
and tens of thousands of specimens, and there are all kinds of
unexpected problems. There are problems with trying to decide
whether names of localities are really different, (Chauntempan,
Chiauntempan, Santa Ana Chauntempan, Tuxpan Ver., Tuxpan Mich.,
Tuxpan Pue.); problems with information as "5km north east of"
means straight line, along a road, along a path? Problems with
inconsistent within-record information, like the label text
specifying the record to be within certain state but the geocoding
turning out to be outside it, or in the border. And what is the
border? Remember this is a fractal concept, depending on the
scale. The precise location of the border between two states
depends on the scale of the maps you use. Problems of inconsistent
among-fields information, for example, the altitude reported in the
label inconsistent with the corresponding altitude from the
digital elevation model, or inconsistent with the ecoregion your
GIS believes the locality belongs to, etc...
In CONABIO we have a team working on the above problems, trying
to conceptualize them as well as to provide algorithms and
practical solutions. I believe this a very interesting and
underdeveloped subject. Many people are now georeferencing label
information, which in our experience is worth all the effort,
because the interface with the GIS allows modelling, extrapolation
and all sorts of interesting analysis. However, the exercise is
fraught with methodological (and some conceptual) problems and I
would love to hear about some of the experiences of you people out
there.
From: Greg.Chandler at anu.edu.au (Greg Chandler)
Hello I read your message on TAXACOM, and can at least supply you
with some sort of feedback. I worked for a while (c. 2 yrs)
databasing the collection at the Australian National Herbarium and
the Australian National Botanic Gardens. It gave me a lot of
experience in lat/longing. I'll address the biggest problem first,
that of augmenting imprecise localities with precise lat/longs.
This is very difficult, and for most of them it is impossible, and
better to give no lat/long than a dubious one, eg. I struck a
number of specimens marked as NW Australia or Central Australia. I
some cases (rare) I could go and look up diaries for collectors and
get some sort of idea where they were, but the lat/long was still
imprecise. What we do at the herbarium is to give it a precision
code (eg. 4 is within a radius of 25 km, 3 is within 10 km, etc).
Then, when you want to pull out localities for precise purposes
(eg. GIS), you can ignore certain precision codes (which still
makes your data mappable, because you ignore the more dubious
localities). So augmenting dubious localities is difficult, and
can be misleading.
Other than that, lat/longs are very useful. I find this now that I
am a research assistant, if collections have lat/longs it is
quicker and easier to look them up on a map, for example, and for
creating maps (using computer softwear). Anyway, there are my
comments. Not many, but it is a worthwhile exercise, and good luck!
________________________________________________________________
Greg Chandler
Division of Botany and Zoology
School of Life Sciences
Australian National University
Canberra, ACT, 0200
Ph. 06 249 3828 Fax. 06 249 5573
From: "Thomas E. Yancey" <tey7004 at geopsun.tamu.edu>
The topic of recording geographic information and of recapturing
this type of info on older collections was discussed in this list
about a year ago. Look up the Taxacom archive and check those
messages for good discussions and insights. I would add that lat
& long is good for purposes of location, but that placing the
collections in a UTM metric grid location is better. I do both
types of determination with our own collections and find it much
easier to do UTM grid locations. Off the shelf computer programs
that use geographic data generally employ lat & long, but you
should consider the need for accuracy and greater ease of use that
UTM grid offers and not let that constrain you. If you get into the
matter of recapturing locality data, be cautious and accept the
fact that real accuracy is impossible, unless the locality can be
physically relocated for checking.
Tom Yancey
Dept. Geology Texas A&M University
From: Jeff Waldon <fwiexchg at VT.EDU>
It is possible to geocode place names using the USGS gazetteer
data, however, our experience with museum collections (especially
ones prior to 1960) have not been very good. The problem of
overstating accuracy is real, but more so than that, you may very
well have problems with the recorded location on the museum tag not
being the place where it was collected, but rather where the tag
was filled out, e.g., at the train station, nearest town, etc.
Quite a few place names don't appear in the USGS gazetteer
especially towns, and in some cases multiple occurrences of town
names can be confusing (how many Pleasant Valleys does CA have?
Virginia has several.) Given the problems inherent in this, we
have still used museum collections extensively, but only for
fairly large area presence/absence coding like hydrologic unit or
county and only after verfication by someone that has a good
handle on the species in question. We've also played around with
accuracy modifiers for distribution, but that runs the risk that
users won't pay attention and get into problems because of it.
Jeff Waldon, Project Leader
Fish and Wildlife Information Exchange
Dept. of Fisheries and Wildlife Sciences, Virginia Tech
205B Washington St.
Blacksburg, VA 24060
(540) 231-7348 voice,
(540) 231-7019 fax,
fwiexchg at vt.edu email
From: James Lyons-Weiler <weiler at ers.unr.edu>
Your concern for precision is correct, however, IMHO it is a matter
of scale. If you will be inferring changes in species
distributions at a scale that is gross (continental or regional)
lat/long should suffice. Also, stating your concerns in the
publication, and making explicit the consequences of imprecision on
the validity of your inferences should satisfy. If you can
determine that the inferences don't change when you make them with
your "imprecise" locations alone (as compared to more precise
locations, if you have them), you've shown that the data are good
enough.
Hope my perspective helps.
James
From: Suzanne McLaren <mclarens at clpgh.org>
I work on the Committee on Information Retrieval of the American
Society of Mammalogists. We have had a set of documentation
standards in print since 1979. However, we recently have
published an update which includes some suggested new fields for
collection databases. As with the previous edition, some fields
are considered "essential" to have while others ar called
"preferred," and there is a third tier called "optional." One of
the newly added "optional" fields is called "Coordinate Precision
Index." It is used to indicate the reliability of coordinates
that have been applied to a collecting locality. One designation
indicates lat/long provided by the original collector using GIS
technology; several other designations are for collectors using
maps of different scale. However, there are further layers of
designations for localities derived from look-up tables, etc. I
think the idea fits very well for "value added" localities. I think
this idea could address your concerns by providing a mechanism by
which users of the data could be made aware of your concerns about
accuracy but without using a huge amount of space.
Sue
Suzanne B. McLaren
Collection Manager, Section of Mammals
Carnegie Museum of Natural History
Edward O'Neil Research Center
5800 Baum Blvd
Pittsburgh PA 15206-3706 phone 412-665-2615; FAX: 412-665-2751
From: "Patricia Jennings" <JENNINGP at cncpe.ecape.gov.za>
We have a possibly related problem, in that on our reserves we need
to map locations of infrastructure (precise) and observations of
animals (imprecise). O.K. so we don't map them on the same layer,
but somewhere/somehow the user must be informed of the
accuracy/precision of each layer = metadata file. Would it not be
some way towards a solution if you were to split your data into
layers of increasing precision, such that your metadata file could
classify them according to their precision. For example layer 1
province level, layer 2 magisterial district, layer 3 within 50 km
radius, etc. Or is this far too simplistic a solution?
Cheers
Pat
Patricia Jennings
jenningp at cncpe.ecape.gov.za
Tel: 0027-41-3902127
Fax: 0027-41-337468
East Cape Nature Conservation
P. Bag X1126
Port Elizabeth
6000
Republic of South Africa
From: Metzler <spruance at infinet.com>
You might want to consider a scale of relative uncertainty
(fuzziness scale) as described below. This was also printed in the
ACS newsletter a couple years ago
Eric H. Metzler
Ohio Department of Natural Resources
614 265 6501
spruance at infinet.com
THE COMPREHENSIVE SURVEY OF OHIO MOTHS AND BUTTERFLIES
completed by The Ohio Lepidopterists
Project No. NGSCW-91-11
Information pertinent to the concept of RELATIVE CERTAINTY
in the final report to ODNR
30 September 1992
The Ohio Lepidopterists strived to provide detail about the capture
localities of the records in the collective databases. The detail
included all information from the locality labels associated with
each specimen. In addition, The Ohio Lepidopterists sought to
provide the latitudinal and longitudinal coordinates for as many of
the collecting localities as possible. The purpose of this report
is to provide information pertinent the methods used to calculate
those coordinates.
Two thousand one hundred fourteen discrete collecting locations
were regi stered for specimens collected in Ohio. The exactness
of the localities varied from vague, such as someplace within the
state of Ohio, to extremely specific, whereby the stated location
is a few feet. The center of each location was used to record the
latitude and longitude, but given the unequal size of the recorded
locations, a method was developed to stipulate the size of the
geographical area represented by each locality.
A SCALE OF RELATIVE CERTAINTY was conceived to represented the size
of the area for each locality. The Scale of Relative Certainty was
designed to reflect the precision of the data labels. Since some
labels indicate a linear site, i.e. "along highway 82", the Scale
of Relative Certainty has two scales, one for square area, S1, S2,
etc., and one for linear sites, L1, L2, etc. The Scale of Relative
Certainty is printed at the end of the report.
The method of plotting Latitude and Longitude is straight forward.
If the collector knew the latitude and longitude, those
calculations were used. For most other sites, county maps as
published by the Ohio Department of Transportation (ODOT) were
used. Occassionally the ODOT maps were supplanted by 7.5'
topographic maps, other county maps, or other city maps.
The ODOT county maps included tick marks for latitude and
longitude. In a few cases, i.e. Lake County and Ottawa County, the
tick marks were found to be in error, thus The Ohio Lepidopterists
used topographic maps to replace the ODOT tick marks with correct
coordinates.
The locations written on the specimen data/locality labels were
located on the county maps, and a rule was used to calculate the
latitude and longitude to degrees, minutes, and tenths of minutes.
For locations that indicated an area, i.e. Vinton County, Brown
Township, Section 11, the center of the area was used as the
precise spot for the latitude and longitude. For linear sites, the
exact center of the linear site was used to calculate the latitude
and longitude. The Scale of Relative Certainty was employed to
determine the correct uncertainty code.
The uncertainty codes were applied starting from largest area to
smallest area. Localities were not made to fit into the most
precise code. For example, the square area of the City of Columbus
is larger than a township but smaller than a county. Rather than
make the square area of the City of Columbus more precise than
stated on the label, the code for county, S7, was used. Given the
diversity of habitats in an area the size of Columbus, accuracy
greater than what is stated on the data label creates precision
that is unwarranted. As the sizes of the localities get larger,
useful detail about the habitat requirements of the species are
lost. Future researchers can use the database to access the
specimen if they need to know more about the sample.
The codes of relative certainty were recorded in the database in
degrees, minutes, and tenths of minutes. The computer used these
data to calculate the coordinates in degrees, tenths, hundredths,
and thousandths of degrees.
The Scale of Relative Certainty follows:
SCALE OF RELATIVE CERTAINTY
September 28, 1992
CODE DEFINITION
S1 Precise location, plus or minus 200 ', is known.
S2 Location is known to be within a circle 1/4 mile in
diameter.
S3 Location is known to be within a circle 1/2 mile in
diameter.
S4 Location is known to be within a Section or
equivalent.
S5 Location is known to be within a circle 2 miles in
diameter.
S6 Location is known within one township (about 36 sq.
miles) or equivalent.
S7 Location is known to be within a County.
S8 Location is known to be within 1/2 of the State of
Ohio.
S9 Location is known to be within the State of Ohio.
L1 The linear site is known to be no more than 1/2 mile
long.
L2 The linear site is known to be no more than 1 mile
long.
L3 The linear site is known to be no more than 2 miles
long.
L4 The linear site is known to be no more than 6 miles
long.
L5 The linear site is known to be no more than 36 miles
long.
L6 The linear site does not exceed a distance equal to
1/2 of the State of Ohio.
L7 The linear site does not exceed a distance equal to
the diameter of the State of Ohio.
From: Una Smith <una at DOLIOLUM.BIOLOGY.YALE.EDU>
In order to relocate a number of old (1800's) localities, I started
by plugging the locality names and what little other information I
had at the time into an online geographic names server. That was
good enough to narrow my search to one or two local area maps. I
used the maps to determine more precise AND more accurate
coordinates than were obtained from the GNS.
Use of a GNS is fast and easy to automate. But I think it is
important to annotate any records containing such data so that the
provenance of the coordinates can be retraced and, if necessary,
the coodinates can be corrected or improved in the future.
The SMASCH <http://ucjeps.herb.berkeley.edu/smasch/index.html> data
model tries to capture the provenance of individual data points
such as coordinates.
Una Smith
Department of Biology
Yale University
From: "Lawrence F. Gall" <lfg at george.peabody.yale.edu>
Thought I'd reply off-list and say we have some experience here at
Peabody with the GNIS, and are mounting an effort to backfill
lat/lon data (with simplistic precision attribution) into some of
our datasets for older collections using GNIS town/place names
(which is very often the only entity in question available from
the specimens).
No doubt you've already seen the USGS' GNIS web offerings at
http://www-nmd-usgs.gov/www/gnis. Peabody has also kept a copy of
the GNIS online since 1994ish at gopher://gopher.peabody.yale.edu
(it'll be migrating into an html context in June of this year).
Happy data-ing...
Larry
Lawrence F. Gall internet: lawrence.gall at yale.edu
Systems Office voice: 203-432-9892
Peabody Museum, Yale University FAX: 203-432-9816
New Haven, CT 06520-8118 USA
From: Arthur Chapman <arthur at erin.gov.au>
There are a number of programs developed in this way - I am not
sure how successful each has been.
The most promising seemed to be one developed under a National
Science Foundation grant by Matthew McGranaghan from the University
of Hawaii. This was very complicated as they wrote their own GIS
for it. Basically, you typed in your locality and a map would pop
up on the screen and then you could grab the point with the mouse
and move it ("I was actually on the road"), etc. We were given an
early draft version some 18 months or so ago to evaluate, however
we have never had the time to do so - there were problems with
loading Australian maps into the system.
matt at uhunix.uhcc.Hawaii.edu. One of the other researchers was a
graduate student called Nayak (nayak at uhics.ics.Hawaii.Edu)
A second was one that we had developed, however it was a very
"buggy" program and we have never got around to fixing it. I had
heard someone else was attempting to, but not heard how this went.
(I have just emailed him to check, I will let you know if there is
anything favourable to report).
A third was one I have never seen - it was a PC program developed
in Canada that I heard about from Larry Spears at the Canadian
Department of Agriculture (speersli at em.agr.ca). This was quite a
few years ago, and I never found out more from Larry, so perhaps it
wasn't any good.
A fourth is one being used by the Western Australian Herbarium -
for that, they have to have their database set up to have fields
such as "nearest named place" "distance", "direction". This seems
to work for them with their gazeteer including old collection
localities such as "the 485 Mile Peg on Great Northern Highway",
etc.
A fifth possibility further down the line is some thought that when
USOBI gets off the ground thatthey will attempt to develop one.
This will be well down the line, however.
There are some problems in all this, and any attempt should include
a field on accuracy. For example - some collectors, when they say
45 miles N of XXXX, mean 45 miles along the road; others mean 45
miles as the crow flies (sorry, an Australian expression). In each
case there is a degree of error that SHOULD be recorded. Also "N"
would generally be an arc somewhere between NNE and NNW and the
greater the distance from the origin,
the greater the error. Any program that doesn't include such an
error figure would, to my mind, be difficient.
Some of us are looking at the possible use of "Java" on the World
Wide Web to quickly ascertain a lat/long. This shouldn't be too
hard - but again, the error factor is a problem.
Sorry that I cannot be of more help in this, but the quickest
solution would seem to be something along the liens of the WA
Herbarium (4 above) unless the Hawaiian University one works. I
have heard recently that it seemed to be a failure, but I have not
had that directly.
Being NSF funded, it is, of course, public domain and was set up to
use USA maps.
I would appreciate any additional information you may find in your
search, as I am still interested in persuing this.
Sorry I can't be of more help
Regards, arthur
Arthur D. Chapman
ºScientific Coordinator, Biogeographic Information, ERINº
Environmental Resources Information Network
internet: arthur at erin.gov.au
GPO Box 787, Canberra, voice: +61-6-274 1066
ACT 2601, AUSTRALIA fax: +61-6-274 1333
From: Kelly Cassidy <kelly at cqs.washington.edu>
For the Washington State Gap Project, we put the herp data in as
point locations, i.e., as a digital cover. We dealt with the
precision problem by having an attribute for precision for each
point. We had 3 levels: within 1/4 mile, within 1 mile (I think;
vicinity.
The hardest thing about inputting lat/longs is that they are often
difficult to derive from a map and the risk of error is high. You
might consider digitizing the point, then letting the software
determine lat/long.
Kelly Cassidy
From: Blair_Csuti at mail.fws.gov
The project you describe is a very good idea. GIS point
localities of specimen records will have many useful applications.
I suggest you send your inquiry to Dr. Larry Master
(lmaster at tnc.org), chief zoologist for The Nature Conservancy.
Larry can provide you with more details on TNC's past and present
approaches to the same issue. Despite past friction between TNC and
ASC, I believe TNC has a lot to offer on this topic. I come from
a museum background (MVZ) and worked as California zoologist for
TNC for four years, so appreciate the task you are facing,
especially since many of the USNM specimens are old and have very
general locality information. TNC classifies localities into three
catagories, represented on maps by circles of about 1/8th mile, 2
miles, and 10 miles, reflecting precision of the locality
information. If a locality can't be pinned down to plus or minus
5 miles (10 mile diamater), they don't use it. Making the jump
from point localities (which can generate dot distribution maps) to
range maps is more complicated. In a GIS environment, points can
be displayed but are not readily analyzed. You need to convert the
distributions to either a rastor or vector format. My program Gap
Analysis) recommends using the EPA EMAP 635 square km hexagonal
grid system as spatial accounting unit. In cooperation with TNC,
we have consulted state reference works and experts to then assign
a probability of occurrence of the species to hexagons for which
there are no specimens, but which are within the distributional
limits of a species. Many old-school zoologists object to this in
principle, but in reality, the range maps in their own publications
do the same thing: draw a polygon around marginal records and
assume the species is likely to be present in appropriate habitat
within that polygon.
For even more resolution, GAP intersects the hexagonal
representation of species distribution with a vegetation (habitat)
map, selecting polygons of appropriate habitat and rejecting
inappropriate habitat. This brings up a number of scale problems
(e.g., we can't map microhabitats that are important to many
species), but at landscape scales (tens to hundreds of square km)
it provides reasonable predictions of species presence. See
Edwards et al., 1995,Conservation Biology 10:263-270, for a
discussion of accuracy of this approach. If you are interested in
more details, I can send some reprints and manscripts on the topic.
Larry can provide details on TNC's rationale for coding locality
precision as well. Another use of GIS locality display is picking
out extra-limital records as candidates for re-confirmation of
their identification. Often records that come from unusual
localities are simply mis-identified!
Blair Csuti
From: "Panian, Mike" <mpanian at NELSON.env.gov.bc.ca>
Here in BC we are working on consolidating data as well. We
translate utm coordinates from an excel spreadsheet to our GIS
(ArcInfo) quite readily. Historical information about species is
often focused on a watershed and often the names are mispelled et
cetera. In such cases we first error check the data and then
attach the location name to the name of a watershed that exists in
our database under another theme that is readily mapped..our
watershed atlas. Then we can attach sightings directly to the
name and accumulate them accordingly. Still another scenario is
having such things as sightings and hunter stats connected to
artificial managment units.
All of this begs the question of metadata and relaibility in
general. I keep wishing for some mappable indication of confidence!
Blurry lines = lack of confidence?
Mike Panian, Regional Inventory Specialist
Ministry of Environment, Nelson BC
----------------------------------------------------------------
From: Jim Reichel <jreichel at nris.mt.gov>
Easy. Put a code in another field to indicate precision. Heritage
Programs nationally use S= within a 3 second radius; M= within a
minute radius or about a mile; G = General Within 5 miles; U =
unmappable.
For a coverage of herp specimens and observations we also
approximately used this and it worked quite well. We did add one
more catagory of 5-10 miles. Then in GIS you can buffer the point
lat-long with the appropriate precision level and get a circle
which will include the actual (presumably) point of collection.
We've actually done this for USNM amphibians and reptiles (from
MT).
It's not easy to do, even for a geographic area you're familar
with, especially old material with outdated place names.
Cheers!
Jim Reichel
Montana Natural Heritage Program
1515 East Sixth Avenue
PO Box 201800
Helena, MT 59620-1800
(406) 444-2546
FAX (406) 444-0581
home page: http://nris.msl.mt.gov/
From: Gary Waggoner
The question of geocoding is difficult because I presume you
have (at least) some pretty reliable locations, some medium
precision ones and then some shot in the dark ones. Therefore, you
don't want to overstate precision on all points but by the same
token you don't want to degrade all points either. I would advise
"classifying" all your points into X (3-5) number of bins depending
on your perception of the degree of precision.
This would be kept as an attribute (or additional data
field) of each point. The best approach would be to probably do
an individual, independent estimate of the potential error for each
point, e.g., +/- 100 meters, +/- 10 kilometers, and enter that
estimate in a data field of that particular specimen record. After
entering all points, look at the data and see if you can lump
error/precision into classes.
Gary
From: jennings at uidaho.edu (Mike Jennings)
The Gap Analysis state project vertebrate specialists have spent
quite a bit of time thinking about this problem. For starters, see
our home page at http://www.nr.usu.edu/gap, especially the "How To"
manual section (you may find the entire vertebrate distribution
modeling section of interest), and the general vertebrate section.
There should be a short article in there by Masters and Jennings on
using 635 sq. km. hexagons to tesselate the land area for recording
specimen localities (smaller hexagons can be used too, and a
georegistered grid is available; let me know if you do not find the
article; also see the 1994-1995 GAP Status Report). We found the
hexegon configuration works best because it fits with the curve of
the earth, like the hexegons on a soccer ball -you cannot do that
with squares.
I will post your query in the GAP vertebrate area of our home page.
You may also want to contact some of the GAP state project folks
directly, as they have done / are doing a lot of plotting of
collection records on a per-state basis.
You may also want to ask advise from Larry Master, TNC vertebrate
ecologist: lmaster at tnc.org (Larry is at the TNC Northeastern
Regional Office in Boston).
Good luck,
Michael D. Jennings
Gap Analysis Program National Coordinator
National Biological Service
530 S. Asbury, Suite 1
Moscow, Idaho 83843
voice: 208-885-3555
fax: 208-885-3618
From: "James K. Jarvie" <jjarvie at OEB.HARVARD.EDU>
Just dealt with this issue here. We have about 3,500 specimens
databased from Borneo, collected with GPS readings by Alison Church
and myself. We also have about 3,500 tree and shrub historical
collections databased from the herbarium in Bogor. These labels
varied greatly in detail. Data are held in MS-Access and we've just
managed to map them in ArcView.
What I've done is to extract and send the Church/Jarvie collections
as one file to ArcView, mapped as a red-dotted theme. Thus red
dots denote accurate locations.
Historical collections, without lat/long data, have been linked to
a file called "Nearest town". This is a list of the populated
areas of Kalimantan with lats and longs (the original data were
downloaded from the Defense Mapping Agency). The tables are linked
and a file is output to ArcView as another theme, mapped with black
dots. Thus black dots indicate approximate locations.
The last theme contains points of varying accuracy; state capitals
and villages are "equal" on the resulting maps. We are looking
into ways of enhancing the accuracy conveyed here.
Bottom line is that data from a single Access database are used to
createdifferent themes in ArcView, with the themes denoting
different levels of accuracy.
Hope this helps,
Jim
James K. Jarvie, Ph.D.
The Arnold Arboretum of Harvard University
22 Divinity Avenue
Cambridge
MA 02138
Tel: (617) 495 3260 Fax: (617) 496 6460
E-mail: jjarvie at oeb.harvard.edu
From: Stan Blum <sblum at violet.berkeley.edu>
With regard to strategies for recording "accruacy/precision": it
seems that a lot of GIS people just recommend that lat-Long be
expressed in decimal degrees and that precision simply be a
function of how many significant digits you record; i.e., -89.2
degrees latitude implies somewhere between -89.15 and -89.25; where
as -89.2146 implies somewhere between -89.21455 and -89.21465. The
problem I see with this approach is that in the computer database
world numeric data types cannot distinguish between 89, 89.0, and
89.000000. Anytime a decimal number ends in a zero digit, this
strategy for handling precision breaks down; i.e., the number can't
be distinguished from one that has any more zeroes appended to
express a greater precision.
The bottom line: I have yet to hear of a strategy that is
universal.
Cheers,
Stan
-----------------------------------------------------------------
Stanley D. Blum e-mail: sblum at violet.berkeley.edu
Museum of Vertebrate Zoology Office phone 510/643-0352
3101 Valley Life Sciences Building
Univ of Calif, Berkeley MVZ front desk: 510/642-3567
Berkeley, CA 94720 MVZ Fax: 510/643-8238
From: WCS Gallon Jug <galljug at btl.net>
We have developed the Belize Biodiversity Information System which
has linked both Advanced Revelation Textual Data to our GIS. We
are working on re-patriating all museum data gathered from Belize
in accordance with the Treaty on biodiversity.
We have mammal data and bird data from many major museums as well
as the NEODAT freshwater fish data from Phil Acad. Sci. I would
like to get data from specimens collected in Belize. The ROM has
provided a great amount of data and what I am doing is checking all
locations in the gazetteer database we are building and once data
are entered here I update locality data by species into our GIS and
this is providing species by species overlays for Belize. What I
do for the museums is return a copy of the data with suggested
corrections/ additions to localities so it can be correctly
geo-referenced when data permit. If desired a Lat/Long data field
can be added to each record when data permit. If you have digital
catalog data for vertebrates collected in Belize I would like to
incorporate the collections as well and the locality data can be
returned with corrections. We have seen many erroneous locations
due to transliteration of place names from Spanish/King's
English/local Creole English/Maya etc. etc. We have most place
names geo-referenced and as early collectors where limited to the
few roads in the country it is not too difficult to figure out
where they really collected. This applies to spelling errors such
as Chan Chin River which is really Chan Chich Creek, the old maps
had errors as well, but one can not find Chan Chin River on any
map. Let me know how/if you would like to proceed and at least for
a small subset of the data you can have help.
Cheers from Gallon Jug
B.W. Miller & C. M. Miller
Tropical Forest Planning Project, Belize
Wildlife Conservation Society
Founded 1895 as New York Zoological Society
Mail: P.O. Box 37, Belize City, Belize, Central America
Ph/Fax/Ans Mach. country codeº501º-21-2002
More information about the Taxacom
mailing list