Hi all.
We have some doubts about the cause of the differences in DN between tiles corresponding to the same scene in Sentinel-2:
S2A_MSIL2A_20240103T160651_N0510_R097_T16PFS_20240103T212452.SAFE
S2A_MSIL2A_20240103T160651_N0510_R097_T16PFT_20240103T212452.SAFE
We have found that the digital level of a shared pixel of the common area in both images is different, and we don’t know the cause of it (angle of solar incidence, etc). I would be very grateful if someone could tell us if some kind of radiometric (or other) normalization is necessary to mosaic both images.
Best regards and thanks in advance.
The values can be different due to the applied AC. It uses some tile dependent value. Thus, the different values in the overlapping area.
1 Like
The difference is usually not that big. You are looking at 2 different pixels here - different values of X and Y in your screenshot. The differences are much smaller in this case - 3rd decimal place, which should not have a big impact on most applications.
for a bounding box of [646480.0, 1191290.0, 646510.0, 1191300.0]
The corresponding numbers are
[0.2364, 0.2828, 0.2036]
[0.2406, 0.2874, 0.205]
1 Like
Thanks @piyushrpt
Yes indeed; the pixels of those images don’t match. Subsequently, I made the comparison for the same pixel and the variability is reduced. I don’t know if this can be useful to anyone, but I did some basic statistics of the average variability in the reflectance (for tropical dry forests and grasslands) and these are the results (bands 4 and 8):
| Band 4 |
Media |
Median |
StDeviat |
Varianz |
| Dry forest |
0,0009 |
0,001 |
0,0002 |
0,00000007 |
| Wet Forest |
0,0074 |
0,0068 |
0,0026 |
0,0000072 |
| Grasslands |
0,0001 |
0,00008 |
0,0001 |
0,000000012 |
| Medium |
0,001 |
0,00096 |
0,00139 |
0,0000019 |
| Band 8 |
Media |
Median |
StDeviat |
Varianz |
| Dry forest |
0,00085 |
0,0004 |
0,001 |
0,000001 |
| Wet Forest |
0,042 |
0,02 |
0,05 |
0,002 |
| Grasslands |
0,00088 |
0,0005 |
0,00084 |
0,0000007 |
| Medium |
0,0023 |
0,0005 |
0,0129 |
0,000168 |
Band 8 (left) and Band 4 (right)
open for discussion 
