The map is based upon free ESRI shapefile vectors hosted here:
http://nationalatlas.gov/atlasftp.html#geology
for U.S. conterminous and the rest;
http://geogratis.gc.ca/geogratis/en/option/select.do;jsessionid=2B48BF4C258172E4AED10E6566E85417?id=E9627B13-9ED3-4100-EEA1-16817F1F36FC
for Canada (despite its title, the geo ages are within the DBF)
http://certmapper.cr.usgs.gov/rooms/we/index.jsp
for the rest of the world but the western Mexico. In the latter case I don't know where to find one, so I hold for good the one from here (after vectorized):
http://www.lib.utexas.edu/maps/mexico.html
This is all of the Earth's geochrons maps I got for free.
The open source GIS software with which I deal with shapes are
http://www.mapwindow.org/ and Quantum GIS http://www.qgis.org/, but mostly the former.
Shapes have several projections, thus all are first converted in geographic projection (WGS 1984, ellipsoid) with Mapwindow, and then overlaid on its long/lat reference grid. Since the maximum graphic output's resolution is 8k, such grid have been split in four part in order to get, at least for VT, 16k; each part has been processed onto GIMP with the antialias filter and re-joined. The 16k map successively has been tiled with the F-TexTools:
http://www.celestialmatters.org/?q=node/77
Now, like I said, the success of the operations above are the roots to assert that the geochronology map is feasible. The problems are others:
1) to uniform the age-colors;
2) the CMOD models;
Hiterto some screenshots of the partial map:
Limit of level 3. For solving the first problem I adopt the color convention of US map for sedimentary, metamorphic, volcanic and intrusive ROCKS. Being the maps diversely colored amongst them, the US dataset is the only dataset whose DBF records shows the RGB colors values. The Canada's map is very detailled and exaustive but doesn't show colors. Bear in mind that Mapwindows doesn't import the ESRI exchange file .e00 and Qgis often fails in read it correctly. Nonetheless, I do manually assign them to the subshapes.
level 4. By the manual assignment of color to the subshapes (shapes made from each eon/era/epoch/period) is possible the export also them as image for being successively "patternized" (I do not like the font choosed for "v" of volcanic. I may find another more sharpen. Instead the "+" of intrusives in near to be fine).
Naturally, level 5 should be still better and in theory would be possible with yet another subdivision of the lat/long grid (4x4) and then to work on 16 pieces of the map... As experiment, on my machine GIMP does riassemble 16 x 8k pieces (32k) of black background in about 30 min. However, Photoshop isn't able to handle 32k images at all. Too bad for such an expensive software.
For what concern the "CMOD problem", is that one CMOD model only is BIGGER, even in binary, and the possible solution should be that of splitting the model in pieces, one for each continents; with the relevant scripts to make it active. CMOD models are yielded through an enhanced version of the Perl script shipped with CMODtracer and fitted to the Zonums http://www.zonums.com/shp2text.html shp2text converter.
So far, just an introduction.
but yes, I know several links hosted. (paywall, though). The map (as well as the others) are here: 
there are the options for datasets not directly linkables from here).
