I try to compute 2 index, REDSI and REHBI (for predicting wheat disease), and in their equations there are different band but also the “central wavelength” of a band.
For example the equations for REHBI are :
((842-665)X(B7-B4)-(783-665)X(B8XB4)) /2
“665, 783 and 842 are central wavelength
of the red band, red-edge 3 band, and NIR band, respectively”
I know how to get the values of the band but don’t understand what mean the values of the central wavelength.
Can someone explain me how to get this values, or whats means this values ?
Thanks in advance ,
Simon.
PS : the 2 article are :
“New Spectral Index for Detecting Wheat Yellow Rust
Using Sentinel-2 Multispectral Imagery”
and
“A Disease Index for Efficiently Detecting
Wheat Fusarium Head Blight Using
Sentinel-2 Multispectral Imagery”
When you take a look at the table of bands for a sensor (e.g. Sentinel-2), you will see that each band has a wavelength. For instance the Red band on Sentinel-2, B04 is indicated as 664.6 nm (S2A) or 665.0 nm (S2B). This is actually not quite true…
A satellite sensor doesn’t exactly measure at a single wavelength but integrates the incoming electromagnetic (EM) radiation over multiple wavelengths: this range is called a band. There is a nice explanation here. The width of a band (the part of the EM spectrum covered) is a trade-off between being narrow enough to measure accurately phenomena (think of a type of light, such as Red or Blue light) but wide enough to let enough energy reach your sensor to get a clear signal.
Each band has a spectral response, which determines how much each wavelength is contributing to your signal. For Sentinel-2 you can find the data in the technical handbook. The central wavelength defines the wavelength around which your interval of EM spectrum (band) is centered. ESA provide such a table on their website, shown below:
Some sensors have a lot of narrow bands (hyperspectral), others cover a wide part of the EM spectrum (broadband). In your case, I imagine that the equation means that you need to fetch the closest band to wavelength:
842 nm in your example would correspond to Band 8 on Sentinel-2.
My problem here is that if i remplace what they call “the central values of wavelength” by the band values, I got 0 …
so either they are lying in the paper or there is another method that they din’t explain.
If I multiply the band value by the coefficient this function gives me for a particular “spectrum” nm (sorry I can’t find the right word for a particular part of the spectrum…) ,
it gives me the value of for the length in question (example: 665) ?
For example :
line 367 of the S2A spectral reponse (in the document) they gave me a coefficient of 0.81 for the band 4/ “spectrum” 665nm.
s if I compute 0.81*B4 ,this gives me the value of the central wavelength or not at all…?
I am not familiar with the methods in the papers that you cite. It would probably be a good idea if you contact the authors to ask them for more information about their approach: in my experience, authors are generally quite open to discussing their science.
