Due to memory constraints, I have a large S1 graph that I must process in two stages. First stage output is the input for the second stage, whereas output of stage 2 is expected to be identical to that of my original graph.
The graph does the following:
The graph is splitted after calibration. This is the first stage:
This is the second stage:
- Write nodes use BEAM-DIMAP.
- First Read node takes a S1 product. The rest load BEAM-DIMAP.
The issue is that when I extract band data from the resulting product, they are close but not exactly the same. There is some variation, although everything remains the same except for the read and save nodes. Any idea why did might happen?