4

I have 3 modis tiles that I want to merge together. They were all reprojected in the same CRS (from sinusoidal to WGS84 UTM zone 32 N), but I cannot succeed. I have tried with merge and virtual raster, but as a result, I get a mismatched map.mismatched map I tried also the GRASS tool r.patch: in this way the tiles match, but the pixel values are different with respect to the original rasters. I cannot figure out what I'm doing wrong...

  • 1
    Perhaps they are not mismatched but the nodata areas are just not made transparent. – user30184 Nov 26 '15 at 15:26
  • there are no nodata values between the 3 tiles... by the way I tried with the SAGA tool "mosaick raster layer" and it seems it worked fine in the end! Anyhow I don't understand why algorithms with the same function give completely different results.. – okapina Nov 27 '15 at 9:13
  • Perhaps not between the tiles but rather inside the tiles. Images in the tiles are warped and rotated and they must contain nodata-areas. For example, if you use gdalbuildvrt you may need to use one or more of the nodata parameters gdal.org/gdalbuildvrt.html. But it is almost impossible to say what you did right or wrong without seeing the exact commands and also sample images may be needed. – user30184 Nov 27 '15 at 11:38
3

In your image it's clearly observed the corresponding area to Italy and its country neighbors. So, I downloaded the equivalent tiles (h18v04, h19v04 and h19v05) to this zone from this Modis product:

ftp://ftp.ntsg.umt.edu/pub/MODIS/NTSG_Products/MOD16/MOD16A3.105_MERRAGMAO/Y2000/

and I loaded the layer with 0 ID in each case. I also loaded the UTM Zone Boundaries and the result was:

enter image description here

There are up to six UTM North zones (31, 32, 33, 34, 35 and 36) in this area. For this reason you cannot reproject in the same CRS (from sinusoidal to WGS84 UTM zone 32 N) each tile before merge them. You have to choose another approach. In QGIS, to merge raster: Raster -> Miscellaneous -> Merge.

At the next image you can observe that my three tiles match adequately after merge process (projection was sinusoidal):

enter image description here

After reprojection to WGS84 long/lat (EPSG:4326):

enter image description here

After 'on the fly' CRS transformation (WGS84 UTM zone 32 N):

enter image description here

  • I see no reason why you could not reproject all tiles to the same UTM zone, as long as they are not too far away (45° or more) from it. – AndreJ Nov 30 '15 at 6:54
  • I edited my answer to include more information. – xunilk Nov 30 '15 at 11:04
-1

That's a very brute force Version, but have you tried to polygonize them separately (computationally a bit of a burden of course) and then merge the .shp files and see if it works?

  • Unfortunately they are continuous raster layers, therefore the polygonization process will completely mess up the values – okapina Nov 26 '15 at 14:05

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.