Hi, Clay.

Your details are daunting, but I'll try to help :)

Since your map is lucky enough to have them, I'd suggest using the
intersections of the latitude/longitude lines as your control points.
Besides offering nice round numbers, they are evenly spaced across the
map.  Try it first with every 10 degrees, and if that isn't good
enough, go with every 5 degrees.

The projection of the original map is not so important in this
process, since you are using control points.  And the datum really
doesn't matter here: the difference between WGS84 and NAD83 is going
to be less than the width of the lines on a map of this scale.  So I
wouldn't worry too much about the map having been folded, either.

As for relying on open source tools, you should be in good hands with
GDAL, GeoServer, and OpenLayers.

Good luck!

Keith


On Dec 5, 2007 11:40 AM, Clay Redding <[log in to unmask]> wrote:
> Hi all,
>
> This is for any of you GIS specialists out there!  Please help me if
> possible, and feel free to email backchannel since this is somewhat
> off-topic for both listservs.  I'm a bit of a newbie to doing GIS and
> geomatics work.  As part of an effort to enable geographical browse,
> search, and retrieval for some digital projects at LoC, I'm trying to
> georeference a historical map [1] of the continental United States,
> as well as parts of Canada and Mexico.
>
> I plan to serve the map as a WMS layer, then pull it in using the
> OpenLayers JS library inside our site for the end user navigation.  I
> aim to "georectify" the image so that it can be placed over top of a
> more modern map layer along the lines of Yahoo! Map Mixer [2] or
> MetaCarta's Rectifier [3].  At this point I'll merely work with
> raster data, since vector data isn't needed yet.
>
> Here are the requirements this project:
>
> -- I am required to use this historical map as the basis of my work,
> as opposed to something more modern like Google Maps on its own.
>
> -- I have to use open source tools, as I don't have ESRI tools at my
> disposal.  Thus far my toolkit includes: Quantum GIS and uDig, GDAL,
> PROJ.4, PostGIS, GRASS (although I'm not skilled at GRASS), and
> Geoserver with PostGIS for serving the content.
>
> Here's what I (think I) know, based on meeting with a GIS specialist
> here at LC who has since departed for a new job:
>
> -- The map has an Albers Equal Area projection, with NAD83 datum, and
> GRS80 spheroid. (In proj or cs2cs: +proj=aea +lat_1=20 +lat_2=60
> +lat_0=40 +lon_0=-96 +x_0=0 +y_0=0 +ellps=GRS80 +datum=NAD83 +units=m
> +no_defs).  She determined this by opening the map in ArcGIS, and
> it's ESRI Well Known Text (WKT) 102003.
>
> -- The physical map was once folded, and might be throwing off some
> of the measurements.
>
> Using both gdal_translate (on the shell) and Quantum GIS (GUI)
> georeference plugin, I've plotted upward of 40 ground control points
> throughout the USA, Canada, Mexico, and Cuba.  I've found the WGS84
> latlong coordinates on databases such as the GNIS as well as the
> USGS.  US points are plotted to 5 decimal points of accuracy.
> International points have 2 decimal points.  I run these through
> cs2cs to convert latlong WGS84 to the Albers/NAD83 coordinates.
> After warping the image based on these Albers/NAD83 ground control
> points, I'll still end up as much as 50 miles off on my north/south
> results, and around 3-4 miles east/west.
>
> My question primarily has to do with best practices for
> georeferencing a map that covers such a large area like North
> America.    Have I used enough points?  Should I use more?  Or have I
> already committed overkill?  Or, do I even have the right projection
> information?  Would I be better off using some other tool(s)?
>
> Thanks for any help you can provide.  I can provide more info, the
> raster image, etc., if needed.
>
> Clay
>
> [1] http://hdl.loc.gov/loc.gmd/g3301p.ct001842
> [2] http://maps.yahoo.com/mapmixer
> [3] http://labs.metacarta.com/rectifier/