0

[OP edited to change tense and to show correct steps, in italics] QGIS 3.10, Ubuntu 18.04

I was questioning why my map displayed at a +10 degree rotation, but displayed unrotated in the Print Layout.

What I did/do now.

  1. Set Project->Properties->CRS to EPSG:5070 - NAD83 (Conus Albers)

  2. Load 8 NED30m tiles covering the Grand Canyon.

  3. Create a Virtual Raster from those and reloaded the .vrt file.

    3.1 Change the newly reloaded VRT Layer's CRS to EPSG:5070

    3.2 Zoom to that layer--to verify that it is indeed not rotated.

  4. Export the zoomed layer as a Geopkg in EPSG:5070.

  5. Exit QGIS.

  6. Restart QGIS, set Project->Properties->CRS to EPSG:5070.

  7. Load the EPSG:5070-saved Geopkg as a new raster layer.

  8. Verify that its Layer CRS is EPSG:5070.

  9. Expect the layer to display with North UP, which it does, with the addition of steps 3.n and the edition of step 4.

  10. The layer also displays properly in Print Layout.

[Harking back to my 4Y3M's-ago question, Project->ProjectProperties->General->CoordDisplay IS set to "MapUnits" as directed by AndreJ, whom thanks again.]

Q. Could someone explain why this is happening, and perhaps suggest another approach, or PCS, to reach my goal; namely, a large-format print of multi-tile b&w elevations? *The corrected steps result in a GCS layer exported to PCS to display squarely/rectilinearly in a PCS project.

Screenshot showing rotation of layer saved-as/reloaded-as PCS CRS

Improve this question
7
  • 1
    Pics or it didn't happen - aka "if you uploaded a screenshot the issue would be easier to judge". – Erik Jun 24 '20 at 6:16
  • 1
    I'd taken the shots, but failed to attach them. Fixed. – quagmire Jun 24 '20 at 12:33
  • This is perfectly normal, since satellite tracks usually run across the earth from NNE so SSW - and the image is oriented perpencidular to the track, otherwise parts of the track would be wasted. You can make the "black triangles" transparent in the layer properties by adding "0 0 0" to the transparent pixels. – Erik Jun 24 '20 at 12:46
  • I am trying to compose a rectangular area which is made difficult by the rotation, not so much by the triangles. – quagmire Jun 24 '20 at 12:52
  • I'm finding it difficult to believe that satellite tracks are part of the explanation because, I assume, Conus Albers projections, surely?, predated the arrival of QGIS (even GIS?), and, I guess, QGIS would be written to output maps similar to what people already expected of paper maps. I could easily, even laughably, wrong, but if so, we might agree that satellite tracks are an undesirable artifact? – quagmire Jun 24 '20 at 13:00
1

Incidentally, "tilt" may be normal, depending on the coordinate reference system chosen for the projection, and how far from the center one looks.

All projections distort, and in fact, different projections are created to minimize distortion in the area of interest (while distorting areas further away more severely) or by spreading the distortion evenly over the entire area.

My map has 8 tiles spanning 2x8 degrees and when viewed in their entirety, in the projection I had chosen, there is clear curvature along lines of latitude. The tiles on the ends are "tilted" compared to the ones in the middle. My bad luck was to pick one of those end tiles and mistake the tilt for some kind of misprocessing.

[Downvoters: please explain why, or all I can do is guess.]

Improve this answer
-1

Here's the answer. I was missing two seemingly innocuous but apparently critical steps; between steps 3 and 4, insert two lines and edit step 4 as a consequence.

Insert steps 3.1 Set VRT Layer CRS to 5070. 3.2 Zoom to that layer (merely to confirm it is rectilinear with the screen: not rotated) 4.[modif.] Export that CRS-modified layer as a Geopkg in EPSG:5070.

Incidentally, this works with single tiles as well as a VRT set. It also worked with UTM12, EPSG:3742.

Summary: in this instance, exporting a layer which has a Geogr. Coord. System (GCS), like say, WGS 84 or NAD83, to file, using a Projected Coord. Sys. (PCS) WITHOUT first changing the layer's Coord. Ref. Sys. (CRS) to a PCS, results in the re-loaded PCS-based file being loaded in a rotated fashion--mimicking the rotation of the original GCS layer.

I will edit my original question to make the correct steps clearer.

Thanks, Erik.

[I think this is a bug because users assume the offer to "save as PCS" will perform the nec. steps to have the orientation correct.]

Improve this answer
5
  • The intention of reprojecting a raster was never to make it appear correctly aligned with a rectangular grid. If someone follows the procedure suggested in this answer, the relationship between the information the raster provides and its spatial location breaks, and any spatial analysis that follows this procedure is going to be wrong. – Gabriel De Luca Jun 26 '20 at 4:11
  • @Gabriel, consider this quote from the wikipedia article on UTM projection: "Projected meridians and parallels intersect at right angles. " People regularly use tools intended for one purpose to perform another. My goal as stated does not include doing any "spatial analysis that follows this procedure." I would be most happy to have a "Newtonian" solution to my problem, which preserves the relationship, but failing that I have had to resort to an "Edisonian" approach to get a result I can use for my intention. – quagmire Jun 27 '20 at 13:00
  • In a Transverse Mercator projection, meridians and parallels are curve families than intersect ortogonally. They intersect at right angles, but they are not straight lines. If you consider the pixel dimensions to be latitudes and longitudes, and you try to convert them to a linear dimension like meters, the geographical position of the information needs to be preserved or the conversion will be wrong. If that didn't matter to you then it shouldn't matter to you that the pixel dimensions are geographic units. – Gabriel De Luca Jun 27 '20 at 13:17
  • @quagmire Yes, the lat/lon lines meet at right angles in UTM because it's using a conformal map projection, but it doesn't say that the right angles are horizontal and vertical (they are only on the central meridian). – mkennedy Jul 12 '20 at 2:51
  • Gabriel and MKennedy, I see it now, thank you. – quagmire Jul 14 '20 at 2:44

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.