Problem reading NASA/USGS-produced geoTIFFs for Venus data

Q&A
31

I decided to dig into the Venus grid the same way as I did with gmt 1-degree elevation grid.

First column of Venus data:

gdal_translate -srcwin 0 0 1 9388 \
   Venus_Magellan_C3-MDIR_Global_Mosaic_2025m.tif \
   first_column.xyz

Last column of Venus data:

gdal_translate -srcwin 18774 0 1 9388 \
  Venus_Magellan_C3-MDIR_Global_Mosaic_2025m.tif \
  last_column.xyz

Column before last:

gdal_translate -srcwin 18773 0 1 9388 \
   Venus_Magellan_C3-MDIR_Global_Mosaic_2025m.tif \
   before_last_column.xyz

The last column is just filled with zeroes. Column before last is not identical to the first column.
so practically the raster x dimension appears to be 18774. More below.

First row of Venus data:

gdal_translate -srcwin 0 0 18774 1 \
   Venus_Magellan_C3-MDIR_Global_Mosaic_2025m.tif \
   first_row.xyz

Last row of Venus data:

gdal_translate -srcwin 0 9387 18774 1 \
   Venus_Magellan_C3-MDIR_Global_Mosaic_2025m.tif \
   last_row.xyz

Row before last:

gdal_translate -srcwin 0 9386 18774 1 \
   Venus_Magellan_C3-MDIR_Global_Mosaic_2025m.tif \
   before_last_row.xyz

The first row contains data. Last row and the one before last are populated with zeroes. According to preview plots, there is large NaN area around the South pole so many rows of cells with zero values are expected.

Based on raster width of 18774, pixel width is 6051000 * 2 * 3.1415926 / 18774 = 2025.12 meters.
Seems reasonable to assume, based on 2:1 ratio, that possible real size of 2025x2025 m Venus raster is 18774x9387. This raster can be extracted from Venus grid by skipping that last column of zeroes and the last row of zeroes:

gdal_translate -srcwin 0 0 18774 9387 \
   Venus_Magellan_C3-MDIR_Global_Mosaic_2025m.tif \
   Venus_Magellan_C3-MDIR_Global_Mosaic_2025m-18774x9387.tif

gdalwarp can be called to interpolate the 18774x9387 grid to geographical coordinates over the global region. Raster size -ts 18774 9387 must be specified explicitly, otherwise gdalwarp inherits coordinate system of the original 18775x9388 raster:

gdalwarp -ts 18774 9387 -te -180 -90 180 90 \
   -t_srs "+proj=lonlat +R=6051000 +no_defs" \
   Venus_Magellan_C3-MDIR_Global_Mosaic_2025m-18774x9387.tif \
   Venus_Magellan_C3-MDIR_Global_Mosaic_2025m-18774x9387-geo.tif -overwrite

Or a new coordinate system can be assigned in place to the 18774x9387 grid (don’t know if this is the correct approach, but very fast and the plot of the resulting raster looks very similar to the output of gdalwarp):

gdal_edit.py -a_ullr -180 90 180 -90 \
   -a_srs "+proj=lonlat +R=6051000 +no_defs" \
   Venus_Magellan_C3-MDIR_Global_Mosaic_2025m-18774x9387.tif

So yes the grid looks buggy in many ways, as @Joaquim suspected.