I use Sentinel-2 MSI Level 2C product to compute NDVI at a 10m spatial resolution - using the the usual formula:
NDVI = (B08 - B04)/(B08 + B04)
but I notice that for the pasture fields that I am examining the mean B08
pixel values are always 5 times greater than those of the B04
band (or the other 10m resolution bands).
Is this typical for NDVI computations?
If not, do I need to scale the B08
values down to compensate for the fact they are integrated over a much wider spectral bandwidth (140 nm) than the B04
values
(38 nm)? - or is that already factored into the product?
Update: Answering my own question:
I looked at the values of the narrowband B8A_20m
band and since they are also about 5 times as big then I guess such values are typical and the NDVI is correct.