Clipping of a Raster changes Range of Cell Values

Question

I'm facing a strange problem. I created a mosaic of two landsat scenes which contains values from ~ -5 to ~11 (Biomass estimation). As my study area is much smaller then the mosaiced Landsat8 scenes, I want to cut the mosaic with a given forest mask shapefile (containing approx 10000 polygons). I just did it with QGIS 2.8 Wien like in several tutorials explained (Raster -> Extraction -> Clipper). It then starts calculating for quite some time, and in the end I get a raster but with values which doesn't make any sense (they are now ranged from 3.4e30 - 3.8e28) and the resulting raster is completely white. I tried it several times with slightly different options, but still the same.

I just have no idea, why it changes any values. Anyone got an idea, how to simply cut the raster into my shapefile without it changing the values?

I uploaded the raster and the shapefile here, if you want to try it yourself.

http://www.filedropper.com/forest-areadata

http://www.filedropper.com/rastermerged43

edit:

• checked the spatial extend, they are both in the same datum (UTM 34N, WGS 84).
• I also checked the data histogram of the clipped ( but basically white) raster, and tried reclassifying it, but still doesn't make sense (and why would it change the data values of the pixels in the first place?)

Answer 1

Your clipping fails because the raster has the odd nodata value of -3.4E+38. Unfortunately, you can not enter that value in the input form.

So I suggest to use gdalwarp to change the nodata value and clip as well, but to the extent of the polygon layer:

 gdalwarp -overwrite -s_srs EPSG:32634 -dstnodata -10 -q -cutline forest_2013_extent.shp -crop_to_cutline -dstalpha -of GTiff RasterMerged43.tif rasttemp.tif

which is reasonable fast and gets this output:

Values are between -1.37 and 9.505, so nothing is broken.

You can create the extent layer with Vector -> Research Tools -> Polygon from layer extent.

Answer 2

I had a similar problem and I solved it as follows.

1. I noticed that some of my polygons were fake polygons, that is they came from a previous cut of polygons and some of them were only lines, not closed up polygons.

2. I manually selected only the real polygons (closed up polygons) because some of the fake ones were unnoticeable because they where "lines" overlapped with the perimeter of the real polygons.

3. I inverted the selection and then eliminate all the "fake polygons".

4. Then I cut my raster with the "real polygon" layer and I got the correct values for my raster.