At NGA, we are attempting to georeference appropriate pieces of our
collections (maps, charts, photos, and docs) that currently are lacking
coordinates. Some items are obviously more amenable to georeferencing than
others, and even those that are sometimes require us to use rather labor
intensive measures to ensure we get an accurate minimum-bounding rectangle
(MBR). For other non-map items, such as photos, we are attempting to use
center-point coordinates derived from a combination of the USGS and Geonames
databases, matching on unique feature name and country. This is a
hit-or-miss proposition, but it looks to be promising. As for matching item
dates to specific point sets for administrative areas that may have changed
boundaries over time, we haven't gotten that far yet. We have yet to find
ready data sources, so we are in the process of building them ourselves.
An interesting question arises on how you group multi-part countries into
one MBR, especially when trying to apply coordinates to textual items
(travel guides, reports, dictionaries, etc...): Would the MBR for the
United States for instance include Hawaii? Should there be MBR's for CONUS,
CONUS + Alaska, CONUS + Alaska + Hawaii? This issue arises with many
countries with detached areas considered at the highest level to be part of
the country, such as Spain/Balearic Islands and United Kingdom/Northern
Ireland/Shetlands/etc...
Automated insertion however is still in our future for those limited numbers
of items that match stringent criteria for each item type and have an
accurate comprehensive data source for coordinates. We are still in the
process of building those data sources. We have some online resources we
have built for Country MBR's, but we have yet to go further down into the
smaller admin zones (states, provinces, prefectures, counties, etc...). A
place we are looking to is the ESRI MapInfo data files, which contain
polygon sets for all these things, from which we could derive the 4 point
MBR's and 24 point polygon sets (to conform to our Voyager ILS).
I am interested to see where this conversation goes.
v/r,
John W Barry [C]
NGA
Senior Software Engineer
EDC / e-Library Team
Bethesda, MD
(T) 301-227-2103 (F) 301-227-5059
"To understand your enemy, you must walk a mile
in his shoes. Then, if he is still your enemy,
you are a mile away and he has no shoes."
-----Original Message-----
From: Subject Coordinates Discussion List [mailto:[log in to unmask]] On Behalf
Of Mary Larsgaard
Sent: Wednesday, September 07, 2005 1:20 PM
To: [log in to unmask]
Subject: Re: [SUBCOOR] batch insertion of coordinates into catalog records
My very limited information on this point is that it's the batch lookup and
insertion that's difficult, although there are certainly other challenges.
For example, we have a couple
of text files that have a geographic area and then the bounding-box
coordinates:
- states of U.S.
- counties of U.S.
- nations of the world
One would have to make sure date of
data in the cartographic item matched the date of coordinates in the lists
(basically 1980s or thereabouts).
And you'd have to find a way to kick
out, say, cartographic items with pre-1900 data (ok when that's the date of
publication; more difficult by a good bit when there's an, e.g., 20th-cent
date of pub but a pre-1900 date of data. If cataloger didn't use the MARC21
date-of-data field, 045, time period of content, then it's real difficult),
to match correct bounding boxes with political area at a given date.
So in theory, one could group one's
catalog records by the geographic
areas in the subject headings (but you'd have to figure out a way where a
record with multiple geographic-area subject headings could have copies in
each group), and then match against these text files, and then move the
bounding box coordinates appropriate to the record into MARC21 |d,e,f and g.
You'd also have to have a way to separate out those geog-area subject
headings that weren't states, counties, natons - which would probably leave:
- cities; could match those against world gazetteer's cities entries; we'd
probably use the Alexandria Digital Library's Gazetteer
http://middleware.alexandria.ucsb.edu/client/gaz/adl/index.jsp
- other political areas (e.g., provinces of Canada; states of Australia):
I've got a list of coords for Canadian provinces and there are probably
lists out there for other political areas of nation's countries
- non-polit areas, e.g., "Denver-Julesburg Basin"; these would be the most
difficult since there aren't any firm boundaries as there are for political
areas.
Am sure I'm forgetting some of the other "challenges"...
Mary
Archie Warnock wrote:
>Mary Larsgaard wrote:
>
>
>>**yes indeedy. I remember a couple years back, a friend of mine who
>>was cataloging photographs of portions of a city was going to enter
>>coordinates into each record, unwisely mentioned that to someone at
>>one of the utilities, who told her no-no, can't enter coordinates for
>>non-cart.mtls. one experiment we'd like to do here at UCSB is download
>>all of the non-cartmtl records in the library's online catalog, pull
>>out all the ones that have geographic-area subject headings (which
>>MARC21 makes very do-able), and then add coordinates to each record
>>(that's the tricky part), and load records into Alexandria Digital
>>Library catalog.
>>
>>
>
>Somewhere around here I had some code that I'd glommed to compute the
>minimum bounding rectangle (MBR) from a polygonal footprint. That
>simplifies the job to the point where it's relatively easy to do
>spatial searching.
>
>For some time, I've maintained the search engine (Isearch) used on FGDC
>metadata records and it does spatial searches (overlaps only, but
>relevance ranked) on bounding rectangles, although it doesn't currently
>compute the MBR from more complex polygons. It's not real MARC-aware
>either, but that could be fixed.
>
>I'm curious where the coordinates might come from because obviously
>this is a job that ought to be automated. Are there any
>easily-available online gazetteers that will provide coordinates from
>place names which would allow batch-insertion of geographic coordinates?
>
>
>
