Bug report #10145
Projection EPSG 2398 inaccurate
|Affected QGIS version:||2.2.0||Regression?:||No|
|Operating System:||Easy fix?:||No|
|Pull Request or Patch supplied:||No||Resolution:|
|Crashes QGIS or corrupts data:||No||Copied to github as #:||18603|
Projecting data from Gauß-Krüger 4, DHDN/Bessel (EPSG:31468), to Gauß-Krüger 4, Pulkova/Krasowski (EPSG 2398) and vice versa gives me data that seemingly are projected correctly, but there is a difference of 3 m mainly in one direction (resulting EPSG 2398 data are too far in the west). In the Halle region (Sachsen-Anhalt) the result should be roughly a simple 2D-Transformation of +22m (east) and +589m (west). The difference can be told e.g. by comparing world files of orthophotos delivered by the Landesvermessungsamt for both projections. However, in the example provided projecting in QGIS yields +19.13 (to the east) 588.10 (to the north) instead of +21.97 and +589.05. I measured by loading into ArcView (3.2), which doesn't provide on-the-fly projections.
Minor differences (order of magnitude 0.5 m) are also detectable comparing projections of QGIS (GK4 -> UTM 32) to those by ArcGIS (which projects reliably, albeit not necessarily reliably on the fly). I was quite shocked. I thought those reprojections have been tested for a long time already in the Open GIS community. I understand that minor differences of less than a meter depend e.g. on the type of transformation used, but 3 meters with not having a choice of transformation parameter sets seem too much to me.
4494022.074 upper left corner (center of pixel, thus 22.074 - 0.1 and 589.039 +0.1 should be difference of upper left coordinate of projected extent)
#1 Updated by Andre Joost over 6 years ago
- Status changed from Open to Closed
This is not a QGIS issue, rather one of PROJ4, the underlying projection engine and its parameter database.
PROJ and QGIS are programmed to use only one transformation for a specific projection, while Arcgis lets the user choose between several allowed transformations.
For DHDN, there are many transformation paramters used within Germany. The best results can be reached with a ntv2 grid.
For Pulkovo 1942(83), PROJ uses the 3-parameter-projection for Czekoslovakia (EPSG transformation code 15998).
There is a 7-parameter transformation available for East Germany. You have to create a custom CRS for that when you use QGIS:
+proj=tmerc +lat_0=0 +lon_0=12 +k=1 +x_0=4500000 +y_0=0 +ellps=krass +towgs84=24,-123,-94,0.02,-0.25,-0.13,1.1 +units=m +no_defs
and http://trac.osgeo.org/proj/ticket/235 for the enhancement ticket in PROJ4.