0

I have a shapefile of European states (open-access from here: http://tapiquen-sig.jimdo.com) and DEM data downloaded from Copernicus.

  • Europe states vector is in GCS_WGS_1984 (Datum: D_WGS_1984,
    • if projected to ETRS89_ETRS_LAEA datum: D_ETRS_1989)
  • Europe DEM tiles (raster) are in ETRS89_ETRS_LAEA (Datum: D_ETRS_1989)

If I overlay both files in ArcGIS, from large zoom they look ok: enter image description here

(ArcGIS on-the-fly reprojection)

but when I zoom in, I see there is a shift of ~200-400 m (I have not noticed any trend going for example south to north, i.e. increasing or decreasing discrepancy).

Is it my raster, or vector data which is not correctly aligned?

Datums in reprojected data looks the same: D_ETRS_1989

I have plotted the basemap layer, and it seems that my raster (DEM, ETRS89_ETRS_LAEA) is better aligned with satelite imagery basemap.

Here is output using Effect tool, DEM is black. When transparent, it is obvious that aligns to basemap better that vector shp (red).

enter image description here

The reprojection of Europe shp from WGS84 to ETRS89 did not help, neither to open reprojected vector in a new dataframe.

How do I "objectively" assure that vector and raster data will align in all the time? (without for example manual shift?)

My fear is, that I can spot that raster and vector data do not align due to shore and islands edge effect. But what if I will have the same shift in my WGS84 GPS point data, and I will just overlook it because I have now way to verify where the data come from?

  • Did you apply a transformation when allowing them to project on the fly? It looks like the two datasets have different datums. – PolyGeo Sep 12 at 7:27
  • I have used the firt one offered by ArcGIS. There are about 10 different transformation option. How can I decide which one is the best?The datums are the same – maycca Sep 12 at 7:35
4

I think the rule would be to use vector that are digitized for the specific scale you will use, for exemple Natural Earth dataset is available at 1:10m, 1:50m, and 1:110 million scales with varying level of detail, you would choose the one that is closest to your intended scale use.

By looking at your picture I would said your first picture is at a too small scale for your vector outline (to many detail) and the second picture is at a too big scale (your vector are only straight line barely following the ''true'' position)

What you're experiencing is very common (and very frustrating) when trying to mix data from different source/provider, you will experience that even very precise data from different source almost never perfectly fit together....

  • Thank you for your answer and understanding. It is also difficult to decide, if it is a problem that I have caused or it just arises from data nature... – maycca Sep 12 at 11:33

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.