Calculation of vegetation indices with 5-spectral band image taken with a drone


#1

Hello everyone. I have used the sentinel 2 toolbox for calculating vegetation indices with sentinel-2 dataset. However, for my project now, i have taken some images with a drone (both RGB and infrared images). Now i have a 5-band raster with the blue, green, red, red-edge (680 - 800) and NIR (800 - 900) spectral bands. These images are just have raw digital numbers (DNs) and no calibration panes were taken in the field for calibrating and converting DNs to reflectance values which are required for calculating vegetation indices.

I was wondering what information does the sentinel toolbox used to calculate vegetation indices. Are the spectral bands sufficient for calculation? i.e. i can just choose the input bands and results are reliable?

Thanks! any advice on this is very useful!

regards,
jessica


#2

if you know the wavelengths and bandwidths of each band you can enter them in the band properties. I am not sure if they are used by SNAP but as you have only 5 this shouln’t take much time.
If you know the order of your bands you can directly enter them into the operators drop-down menus to define which are to be used.


#3

Actually these are just descriptive attributes of a band and they are not taken into account for computing physical values.
In order to convert from DN to physical values you have to apply a linear transform (using bias and offset) to DNs. Usually, this is taken care of by the specialized product readers (liken the S2MSI Reader for Sentinel-2 products). Otherwise, you have the option to create virtual bands on your DN product having as expression this linear transform.
Planning to use directly the DNs will not yield to reliable results, since the indices values usually depend on the aforementioned bias and offset.
Take for example NDVI = (NIR - RED)/(NIR + RED).
If the reflectance value is obtained from DN via a transform like V = aD + b, the NDVI becomes:
NDVI = (a
Dnir + b - aDred - b) / (aDnir + b + aDred + b) = a(Dnir - Dred) / (a*(Dnir + Dred) + 2b), which obviously need the a and b coefficients, not only the DN values.


#4

Thank you for your reply. This means that I have to calculate reflectance value for each of my bands. Do you have any suggestion to do this from DNs, but I don’t have reflectance panes for calibration…