Hi dkelly
let me first answer question 2 and 3.
2. To validate the results, you need in-situ data to compare with. You might need to tweak the conversion factors for tsm and chl, too.
3. For MSI data it is known that the result can look a bit noisy. (SIIITBX-374)
First one is not so easy to explain. because I don’t know what you mean by unrealistic and what you expectation is.
But I assume the resulting images are not colored as you expect, because of the way SNAP is doing this.
The resulting colors depend on the minimium and the maximum values in the image. 2.5% at the upper and lower end of the histogram are clipped, to eliminate outlier.
But if you have many e.g. bright pixels in the image, then the results might not be as expected.
To overcome this you can clip the value range in the RGB expressions.
In this post I explained it and provided an example rgb profile.
You need to adapt the band names and the limits for your case.