Main question: how can I improve the accuracy of my reprojection?

I am combining my dataset to a previous one, and I have chosen a different CRS: the previous one is in WGS 84 (EPSG:4326) and mine is in WGS 84 / UTM zone 36S (EPSG:32736). This allows me to get more accuracy in measurements, and is also easier as I'm using meters (from what I've understood and experienced using functions in QGIS). Both rawdata were obtained in WGS 84 EPSG:4326, but I reproject mine before doing any work on it. I would prefer keeping on using a mercator as the map unit is in meters and zone 36S is the relevant one for my study area.

My project aims at counting points inside the squares, calculate area of intersecting polygons, calculate dispersion.

  • previous dataset in EPSG:4326: calculation has been done.
  • my dataset in EPSG:32736 / UTM zone 36S: the study area is the same + an additional area. Not analysed yet.

To be able to combine the datasets I performed a reprojection with the older datasets using the option "reproject layer" from the toolbox (QGIS 2.18), but the result is slightly off. This can be seen looking at these 2 grids: enter image description here these grids are made of 1x1 km cells. pink grid reprojected from WGS 84 / EPSG:4326 to UTM zone 36S EPSG:32736, white grid created under WGS 84 / UTM zone 36S EPSG:32736.

I have read a couple of questions dealing with similar topics, but it gets a bit too difficult for me who is fairly new to GIS and have never coded before. What I have come up with as solutions so far:

  • using a different reprojection of the zone 36S, though I'd rather avoid to, and I'm not sure that would make a difference in the accuracy

  • it might simply be that the result cannot be improved as it is a fine scale. In that case I'm wondering whether it is a problem or not that the reprojection is slightly off.

  • I could also just use the EPSG:4326 since my data were primarily in this format, and deal with the conversion from degree to meters but it is not easy and less accurate from what I have notcied so far.
  • I could enable the OTF but from what I have understood, measurements would be wrong since the CRS don't have the same map units.

I hope my explanation is clear enough, I might get this whole problem wrong since i have never worked with projections, so I will be very thankful to any contribution or clarification!

  • In the grid created in WGS 84, the cell dimensions are probably not exactly 1 x 1 km. They are more likely measured in decimal degrees, because that's the unit used by that coordinate reference system. – csk Jul 12 '18 at 18:40
  • 2
    A grid created in one projection is not identical to a grid created in a different projection. The two grids you're displaying are different grids, so of course they don't align. I think you should pick one grid and use the same grid for all your analyses. – csk Jul 12 '18 at 18:44
  • @csk think of turning your comment into an answer. – AndreJ Jul 13 '18 at 6:14
  • thanks, I have managed to create a grid in EPSG:4326 for both study areas. It makes it a bit more complicated to deal with degrees. I guess if I reproject all my data in UTM I will still have the small reprojection error as seen from the pink grid, how could this be corrected for? – Julie Jul 13 '18 at 8:29
  • The difference between the black and pink grids doesn't sound like a reprojection error. It sounds like they actually represent different areas. Are your other layers misaligned when you re-project them? How are you verifying that? Are you comparing point locations to known points on a base layer? Eg, taking a point at the corner of a building, then comparing the location of that point to the visible location of that building on a satellite image. – csk Jul 13 '18 at 14:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.