I am trying to align a map image layer such that the defined coordinates on the map match the coordinates displayed in the lower status bar on the JOSM display.
The map is a USGS 7.5 minute series (topographic) with coordinates specified in NAD27. I have used the USGS conversion page to translate the NAD27 corners to NAD83.
Even with that adjustment, the positioning of the topographic layer does not align with the mapnik layer, nor with items downloaded from the OSM repository (which presumably is identical to mapnik)
There is also a very slight variance being introduced from my attempts at positioning the topo layer, due to pushing the scrolling increments of my mouse to its limit. This is minor compared to the difference I am seeing.
My best estimate of the east west offset (between topo and mapnik) is 2.95 seconds of longitude.
I believe that everything in OSM, including JOSM is based on the WGS84 datum. However, as I understand it, the NAD83 datum is at this point in time insignificantly different (at least smaller than typical GPS errors on cellphones used for gathering much of the OSM data) so I am puzzled by what you are seeing.
In JOSM there is a USGS topo layer you can turn on. In my area the imagery that shows if pretty close to what is in OSM and matches pretty well with MapBox imagery. If you turn on the USGS topo layer how does it line up with your version of the USGS maps?
The USGS tiles align pretty well with mapnik. I am puzzled as well.
One possible solution to this puzzle is USGS CORS. There is a CORS station at a small GA airport, about 15 miles from me. Maybe I should drive over there, and take a number of readings (one each on 4 sides of the CORS pedestal), log them with my GPS mapping program, then come back and see how they compare with where mapnik says those coordinates should be.
I think there may be several problems involved with what I was attempting to do. The first is the map projection. The 7.5 minute quad maps are Polyconic, while JOSM is using Mercator (by default). Presumably the tiles (loaded dynamically) are from a Polyconic map, that has been chopped into smaller tiles. To get better resolution (and avoid the 499 errors I see randomly), I’m trying to use the entire 7.5 quad map. Projection errors, from one small tile to the next, may be minor, and vanish into the diminished resolution. It may be more of an acute problem with the entire quad as a single large tile.
Reading about Polyconic, in the USGS professional paper 1395, I find the following …
So there may be problems that are insurmountable.
Then we go back to the actual images of the topo quad maps. They download from USGS as a PDF file. JOSM has no native ability to use a PDF as a background layer. There is a plugin that looks like it can do this, but on closer inspection what it is trying to do is extract map layer information from the PDF (ways, nodes, etc). That causes me to convert the PDF to something that can be handled as a background layer (PNG). I believe that the PDF to PNG conversion is faithful, and not introducing distortion. The original 7.5 minute series were all issued in NAD27. Somewhere around the mid 1980s, NAD83 began to be introduced, but (from what I can tell) was not used so much to change the base coordinate system, as it was to add cross-hairs to show where the NAD83 compliant corners would be. Maps from pre-1983 did not have these. The particular map that I was having problems with was issued in the 1960s. There is a later version of that map with the cross-hairs. Even then, the issues quoted above may get in the way.
It would be handy if JOSM supported Polyconic. I do not see it in the list of projections. I’m going to spend a few more hours with this, to see if I can find a way to use these maps.
I’m taking a different approach now. Rather than trying to position the topo images to an absolute location, I’m trying to align them to match mapnik. That works reasonably well, with one caveat. The NAD83 cross-hairs do not align with where they are supposed to be (sometimes).
One example - Trenton FL quad sheet, NE corner
correct expected position
cursor over the NAD27 (original) corner mark
cursor over the NAD83 (annotated) corner mark
This may all be due to the original TIGER files, I don’t know. I’m just reporting what I see.