I know this may sound like a silly question, and it is, but I've read lots of questions and tutorials, watched a few videos and still can't properly understand what I should do...

I am trying to work with Rio de Janeiro's datasets at two different administrative levels: neighbourhoods and administrative regions. Their shapefiles are made available by Rio's city hall, but are set in different CRS (i'll link them by the end of the post). The neighbourhoods shapefiles are set in EPSG:31983 and the administrative region in EPSG:29193(29183).

How can I get the shapefiles to align correctly?

When loading both shapefiles, the neighbourhoods CRS is correctly interpreted by QGIS (I should mention that I'm using QGIS 3.2), which sets it as 31983. The administrative regions are not, and QGIS generates a CRS.

From my understanding the next step should be to reproject this layer onto its right CRS, 29193 (or 29183, I don't know, but either way it doesn't come up right). When I do this the shapefiles are almost aligned, but they're not. They are some 15 meters misaligned.

Thinking that I might have understood something wrong, I tried another approach. I changed the administrative region's CRS via "set layer CRS" in the properties to 29123, and after that reprojected it to 31983. Again, with no success.

I've tried basically all combinations with reprojecting the layer and changing the CRS, but none worked. The first approach is the one I thought would be correct, but I ended up getting frustrated and trying several others.

The links I mentioned earlier in the post:

  • 2
    These work fine for me. I started my map with the Google Maps coordinate system of 3857, added each shapefile, and they draw fine. Why they are in different coordinate systems is a good question... when in doubt, re-project to 4326 / WGS84! Commented Jul 10, 2019 at 22:31
  • I tried replicating what you did, but it didn't work. Started a new project, setted the CRS to 4326. Loaded both the shapefiles, reprojected them to 4326, but still no good. Did I do it correctly or messed up in the middle? Cheers
    – dhersz
    Commented Jul 10, 2019 at 22:46
  • I just tried all permutations of setting the project CRS from the datasets and they all work fine... really no idea what is going on with your data. Are you unzipping the files? I'm working right off the zip files... Commented Jul 11, 2019 at 2:37
  • I tried adding the zip files and got the same results. When you say you got it correct, did you zoom in substiantially? Not doubting you by any means, it's just that the 15 meters I mentioned in the question are only visible after some good zooming in, so it might have passed as correct even if it is not. Thank you very much for the patience and support. Cheers
    – dhersz
    Commented Jul 11, 2019 at 3:23
  • Hey sorry I missed this (got busy at work). I see the data shift (thought it was just one was WAYYY off in the first question). I tried all combinations of coordinate systems and datum shifts in the latest version of QGIS and can only conclude that there is something happening with the data and the 2 different coordinate systems being used which tells me that there might be 2 different source systems and 2 different datums being used for those datasets. Not helpful, but I don't think it's a QGIS display issue... is it possible 2 maintainers were being used for those datasets? Commented Jul 15, 2019 at 22:01

2 Answers 2


I downloaded both layers, extracted them and imported them into QGIS 3.8. QGIS offers various transformations you can choose from, most transformations say that the accuracy is 15 meters. The best I could find was transformation #4 (see screenshot). my guess is that QGIS 3.2 does not offer you various transformations and chooses a "default" transformation with a difference of 15 meters.

enter image description here


enter image description here

Using QGIS 3.10.x, I was able to define/reproject the RAs layer closer to the boundaries of your Bairro data. I've included cleaned/projected copies of both the Bairro and RAs data in a geopackage below.

Additionally, to further reduce the roughly 1m error introduced above, I derived a RA dataset from the Bairro geometries to ensure it would be 100% aligned/equivalent, and performed a table join against the original RAs data to ensure attributes were preserved. This is also included in the data.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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