I am new to this forum, hopefully somebody can help me out.
I’m doing a Msc thesis on classification of different grassland types in the Netherlands.
Beside other indices, I would like to use the S2REP vegetation index for discriminating between different grassland classes.
I’m working with Sentinel-2 Level2A BOA reflectance data in SNAP 5.0. Atmospheric correction was performed using Sen2Cor at 10 m resolution without any errors.
I’ve used the S2REP processor with B4, B5, B6 and B7 as input bands as described in the SNAP Help section.
My problem is that the output product gets flag 4, saturation (index value is too high), for almost all pixels in my research area.
What could be causing this?
Thank you for your reply. Indeed, I have read that article. The S2REP VI formula they give is the same as the one that is implemented in SNAP.
705 + 35 * ((((NIR + R)/2) − RE1)/(RE2 − RE1))
or with the bandnrs.: 705 + 35 * ((((B7 + B4)/2) − B5)/(B6 − B5))
So in this case there is no need to adjust the individual band factors, or is there?
Also, for my research area the calculated S2REP value for the grassland areas averages around 728, these pixels still get flag 4. Values of 728 do not seem unrealistic to me for healthy grassland vegetation in spring.
you are right - in this case they are considered as constants. So I would have chosen the same values.
Based on figure 5 your resulted values seem reasonable.
I don’t know why they are still flagged as 4. The module was developed by Dragos Mihailescu from CS Romania. Maybe you could directly ask him.
The threshold values for the flags that are currently used are incorrect. The index values should range between 700 and 730 (or, more relaxed, between 690 and 740). This will be corrected.
