[GIS] Fixing raster with wrong Eastings using QGIS and GDAL

gdalqgisraster

I get an error when trying to warp a tif file in QGIS:

ERROR 1: Too many points (35721 out of 35721) failed to transform,
unable to compute output bounds.

I am using QGIS 1.7.0 Wroclaw with gdal 1.8. I am trying to reproject from ETRS89 to HD72 EOV. There was an error in the data, with the UTM zone added to the eastings. That is, the x coordinates read 34327213.050 when they should read 327213.050 in UTM zone 34N. I have tried changing the .tfw file but this did not help. Then I translated the tif to a .vrt and edited the vrt, removing the rogue 34, and fixing the extent. This also did not help – I still get the same error when I try to warp. In layer properties, this raster now has the same origin, extent and SRS as other similar rasters, but only this one gives the error.

How do I work this out?

A similar question (Setting projection of tif file in QGIS?) was asked 4 months ago, but it was not the exact same problem, and the OP did not explain how the problem was resolved.

Best Answer

Have you tried using gdal_translate to fix the error? Something like this should work:

gdal_translate -a_srs EPSG:32634 -a_ullr <actual coordinates go here> broken.tif fixed.tif

The order of the coordinates should be: left, top, right, bottom.

This should ignore everything that's in the TIFF's metadata and overwrite it with the proper bounds and projection. Also, by default, it'll embed the bounds in the TIFF rather than relying on a .tfw, so don't worry if one isn't created with your new file.

A link to the previous question might be useful, even if its not exactly the same issue.

Related Solutions

QGIS Reprojection – How to Reproject a Raster File with Datum Transformation

Gdalwarp is the tool to reproject, you find it in Qgis under Raster->Projektionen->Transformieren or standalone in OSGEO4W.

Basic command is

gdalwarp -s_srs EPSG:25832 -t_srs "+proj=tmerc +lat_0=0 +lon_0=6 +x_0=2500000 +y_0=0 +k=1.000000 +ellps=bessel +units=m +nadgrids=./BETA2007.gsb +wktext" input.tif output.tif

BETA2007.gsb should be in the same folder, or use absolute path to it.

In Qgis, select the loaded tif as input, and its CRS, and paste the following in the target-CRS:

+proj=tmerc +lat_0=0 +lon_0=6 +x_0=2500000 +y_0=0 +k=1.000000 +ellps=bessel +units=m +nadgrids=D:\path\to\your\BETA2007.gsb +wktext

QGIS – Resolving Black Output in Proximity (Raster Distance) Tool

I downloaded the CLC 2018 raster data to test how the Proximity Tool works with this data.

I was able to produce the following output:

To do this I:

Kept the CLC data in its own projection
Using the Proximity Tool, I set the "List of Pixel Values" to 24 (which from what I could tell is a forest class)
Ran the Tool

The result did come up as appearing to be all black (as you experienced) but this seemes to be how the raster is presented, with no streching of the pixel values. QGIS is trying to draw the full range of pixel values (0 to 21900 in my case), so everything appears to be black.

This can be fixed one of two ways:

Right-click the result and select "Stretch using current extent", this should normalise the view and you should see pixels representing distance; or
Open the results properties, and open the "Symbology" menu. Choose the render type as Singleband Pseudocolor and play around with the settings. Here is mine:

Hit apply and it should present as you wanted it. Just remember the resolution of the image is 100m, so try stick to those intervals. If you want to include more of the classes you can either add them in the Promiximity tool's options or you can Reclass the image identifying those pixel values representing Forest.

UPDATE WITH SOURCE DATA

Here is the data as provided through the WeTransfer link:

The legend shows the forest having a pixel value of 23.

Running the proximity tool I set the "List of pixel values" to 23.

The OUTPUT:

I opened the Layer Properties, and the Symbology Tab. Changed the Render Type to "Singleband pseudocolor" Changed the Mode to "Quantile" Changed the "Classes" to 15 Hit the "Classify" button and then Apply

Here is the result:

I think you may be there with your method, just a little refinement in the presentation.

Best Answer

Related Solutions

QGIS Reprojection – How to Reproject a Raster File with Datum Transformation

QGIS – Resolving Black Output in Proximity (Raster Distance) Tool

Related Question