Mosaicking produces NaN pixel

Optical Toolbox
#1

Hi there,

i ran into a problem, which i couldn’t solve on my own and have not found any solution here in the forum.
I’m working with Sentinel 2 Level 2A imagery and there are pixels in some Bands (B1, B6 and others) which have a value of 0.0, probably because they had less reflectance than detectable. When i view those bands in display those pixels appear as NaN (No-data colour is red), as the default Valid-Pixel Expression is B1.raw > 0.

B1 invalid expression
B1 invalid expression941×566 21.7 KB

When i change that expression to B1.raw >= 0, i see those pixels as 0.00000dl.
B1 valid expression
B1 valid expression995×594 21.5 KB

But after Muti-Size Mosaicking two images with all the bands of both images changed to B1.raw >= 0 (B2raw >= 0,and so on), I still get NaN values after mosaicking in those places.
This leaves me with missing pixels after the classification later on.

#2

what happens when you enter another number in the No-Data Value field, for example 99999, save the product and recalculate the mosaic?

#3

Great, that worked! Thanks.

Edit: That was a premature celebration.The pixels are still NaN values after the mosaicking, i just did not see that because i had ‘no data-values used’ ticked off.
This still leaves me with missing pixels in my classification.

#4

well, i found a way around. I just duplicated all the bands with the BandMath function using the expression:
if B#==0 then 0.00001 else B#
though i feel like this is not an aesthetic solution.

#5

thank you for reporting.