5

I'm trying to georeference some WWII maps available at the University of Texas repository (e.g., example map).

The problem is that these maps refer to an unknown proj system (at least for me) and I can not figure out how to set up a custom reference system.

These maps refer to a "Mediterranean Grid" and "Caucasus Grid" as follow:

enter image description here

Is anybody aware of how to manage these kinds of grid systems and how to create a custom projection (and a graticule) on top of them?

In one of the adjoining map sheets, I found the following information:

enter image description here

It introduces new clues and allows the benefiting of this discussion: Setting proj parameters of old map.

EDIT April 5th:

On the basis of the different answers (thanks to all), I was able to generate the following custom projection:

+proj=lcc +ellps=clrk80 +lat_0=39.50 +lon_0=45 +lat_1=39.50 +x_0=2155500 +y_0=675000 +units=m +no_defs

Then I georeferenced this sheet. I tried either with Helmert or Polynomial or Thin Plate Spline transformation algorithm (with just few differences). Yet, the map appears inaccurately georeferenced. See the following example:

curl -L "http://legacy.lib.utexas.edu/maps/ams/turkey/txu-pclmaps-oclc-10127923-palu-e14.jpg" > ./output.jpg
# add control points
gdal_translate -of GTiff -gcp 751.581 2077.58 1.72e+06 600000 -gcp 6302.39 1037.18 1.81e+06 600000 -gcp 6510.44 5478.53 1.8e+06 530000 -gcp 1461.03 5785.78 1.72e+06 540000 ./output.jpg ./output_gcp.tif
# apply the proj parameters
gdalwarp -r near -tps -co COMPRESS=PACKBITS  -t_srs "+proj=lcc +ellps=clrk80 +lat_0=39.50 +lon_0=45 +lat_1=39.50 +x_0=2155500 +y_0=675000 +units=m +no_defs" ./output_gcp.tif ./output_mm_warped.tif

enter image description here

I'm afraid that, as suggested by @GabrielDeLuca, the missing of datum transformation is a big issue. But I have no idea how to collect this missing piece of the puzzle.

In addition, whereas other sheets of the same series report also the scale factor (see for instance the Levant Grid parameters below), to my knowledge, there are no maps of Caucasus Grid that offer this information. enter image description here

11
  • 1
    What GIS Software are you using?
    – Mapperz
    Apr 2 '20 at 13:33
  • The Levant Grid which is also mentioned in this map but which you cropped is - like the Palestine Grid - based on the Clarke 1880 ellipsoid. Maybe this helps. Have you searched for something like "historic turkish crs"?
    – Erik
    Apr 2 '20 at 13:42
  • Thanks for the comments. 1) I use QGIS 2) Yes, I already searched for some references in secondary literature, but with no luck. As fas as I can understand, these graticules are based on the so-called "Modified British System" (echodelta.net/mbs/eng-overview.php). The page I found, however, focuses on Europe only.
    – Pankus
    Apr 2 '20 at 14:35
  • 1
    Try this US army foreign maps manual archive.org/details/TM5-248/page/n1/mode/2up
    – Hans Erren
    Apr 2 '20 at 20:35
  • 1
    Another interesting document to find the datum used: asprs.org/wp-content/uploads/2018/03/09-16-GD-Turkey.pdf Apr 3 '20 at 3:40
5

According to Mugnier ( https://www.asprs.org/wp-content/uploads/2018/03/09-16-GD-Turkey.pdf) the projection of the original Turkish maps is Bonne with the following proj4 definition:

+proj=bonne +lon_0=28.980783 +lat_1=39.6 +x_0=0 +y_0=0 +a=6378249 +b=6356515 +units=m +no_defs

Note that the Turkish Bonne projection uses an ellipsoid that is a rounded Clarke 1880: Clarke 1880 a=6378249.145 b=6356514.870; Turkish Bonne a=6378249 b=6356515

Sheet E.13 harput-elazig is fitted with the following wld file parameters, with 13 pixel accuracy on affine fit, using the following bonne coordinates for the map corners:

LL: 850000,-100000
UR: 950000, -20000

txu-pclmaps-oclc-10127923-harput-elazig-e13.wld
15.86182586668007843
0
0
-15.8094810492159521
827612.23303510167170316
-11119.60709845979545207

Sheet E.14 Palu is fitted with the following wld file parameters, with 37 pixel accuracy on affine fit, using the following bonne coordinates for the map corners:
LL: 950000,-100000
UR: 1050000, -20000

txu-pclmaps-oclc-10127923-palu-e14.wld
15.97090658853279699
0
0
-15.90264386485743309
942247.13803694711532444
-7340.06815728270703403

Both maps georeferenced in Bonne projection with OpenStreetMap overlay in QGIS
Maps georeferenced in bonne projection with OpenStreetMap overlay

British war overlays:

Brown Caucasus grid:
+proj=lcc +ellps=clrk80 +lat_0=39.50 +lon_0=45 +lat_1=39.50 +x_0=2155500 +y_0=675000 +units=m +no_defs

Blue Mediterranean grid:
+proj=lcc +ellps=bessel +lat_0=39.50 +lon_0=29 +lat_1=39.50 +x_0=900000 +y_0=600000 +units=m +no_defs

Source for Mediterranean grid is the definition on this map: https://earthworks.stanford.edu/catalog/princeton-kw52j965x

4
  • Many thanks for your efforts, Hans. Just one more question. How did you get the Upper Right corner of these maps? Specifically, I cannot figure out how you get the -20000. Finally, I don't know if I correctly georeferenced these maps, however also with Hans' method, there still exists a horizontal error between 500 and 1400 meters in one compares results and background baselayer.
    – Pankus
    Apr 7 '20 at 12:55
  • Hi Pankus, the caucasus grid and the bonne grid are both on a clarke ellipsoid so there is no datum shift needed. Remember that the paper distortion is 37x16=600m in E.14, and that the 200k Turkish source sheet predates the 1924 triangulation. On the 800k triangulated map 1mm=800m! For Clrk80 to wgs84 the nearest datum point is Deir ez Zor, so that is worth a transformation test to WGS84.
    – Hans Erren
    Apr 8 '20 at 5:26
  • A Turkish Bonne map sheet is exactly 100 km wide and 80 km tall, so that explains the exact Bonne coordinates, there may be a fiddle factor needed though, as the 200k maps predate a nationwide triangulation.
    – Hans Erren
    Apr 8 '20 at 5:31
  • Remember also that the Bonne maps are based on the Istanbul meridian and that the exact Greenwich relation varies according to Mugnier, so using a different base longitude Is worth a try.
    – Hans Erren
    Apr 8 '20 at 5:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.