Set one output band's no data value with gdal_merge.py

Question

I'm trying to stack two images with gdal_merge.py -separate so that I can use the NoData values in the second band to mask the first band (was going to use gdal_translate -b 1 -mask 2). The first image (or band 1 after the merge) has no NoData values, while in the second image NoData = 0. However when I use gdal_merge to stack the two images, the NoData for the second image/band is removed and set back to 0.

There is a "-a_nodata output_nodata_value" option in gdal_merge, but I don't want to set the first band's NoData values, just the second. Is this possible?

Or is there another way I can use the second image's NoData values to mask out values in the first without stacking the images via merge? That is my ultimate goal. Maybe I can achieve this via gdal_calc?

Answer 1

As I understand, here is your setup:

• A: raster
• B: raster, same size as A
• nodata_b: nodata value for B; 0 in this instance

Your goal is to create raster C, which is A masked by valid B.

C = A where (B != nodata_b) else nodata_a

The question now concerns what are valid possibilities for nodata_a. This is something that you don't give, so there are a few options. Remember that rasters are required to have a value for each pixel. The nodata tag is just there to help interpret such values within the raster array as having a special property (nodata). You need to determine what nodata_a should be.

If nodata_a is allowed to be 0, this is pretty trivial.

gdal_calc.py -A rasterA.tif -B rasterB.tif --outfile=rasterC.tif \
--calc ="A*(B!=0)" --NoDataValue=0

Here, B!=0 is translated into 0 for False and 1 for True on a pixel-by-pixel basis. So 1*value = value and 0*value = 0.

If nodata_a needs to be a specific value, things become a little more tricky. Let's say 0 is a meaningful value in raster A. Setting nodata_a to 0 would mean that meaningful data is absent in raster C. Maybe there are no meaningful negative values in A (and the datatype of the raster permits negative values). So if nodata_a==-999.

gdal_calc.py -A rasterA.tif -B rasterB.tif --outfile=rasterC.tif \
--calc ="A*(B!=0) + ((B==0)*-999)" --NoDataValue=-999

Here, there are two parts to the calculation. On the left is the same calculation as above -- A where B!=0 else 0. On the right, the result is either 0 for valid pixels or -999 for invalid. Summing the left and right values will result in the correct values of A where valid and -999 where invalid.

You could of course substitute any value you want for -999, as long as that value is guaranteed to not be present in the valid portion of A.