Equivalent value for SAR image as NDVI. Need HELP!

Microwave Toolbox
#1

Hi, I am a beginner in RS. Do anyone know how to create a simple mathematical formula or equation for radar image that equivalent as NDVI value in optical image? my purpose is to identify healthy and unhealthy crop tree as nowadays NDVI is the best indicator to differentiate it. so, how with radar images, backscatter coefficient cannot use to differentiate healthy and unhealthy tree easily.

I have tried with my equation from regression analysis of NDVI and Backscatter Coefficient (as image below), but the result was not good and weird.

ndvi sigma result.png992×264 8.4 KB

so, from the equation y=0.1603x+2.3753, I replace the x value with NDVI value (0.8175). I assume as Y is the predicted radar value that I looking for. In the bandmath i write as;

  1. (b1-0.1603*0.8175)+2.3753
  2. (b1-2.3753)/0.1603
    3)(b1-2.3753)/0.1603*0.8175

** b1 is the sigma image of radar (VV)

y value.png930×179 10.2 KB

**Y values was so weird.

Do anyone have some advice and good suggestion to help me get better result or correct way to achieve my purpose. thanks a lot.

Survey vegetation growth by sentinel-1
#2

First of all, I would recommend you to be sure that you are using backscatter values in the linear units. dBs are tricky, so it’s better not to work with them than possible.
Secondly, it would be better to use two polarizations instead of one. For Sentinel-1 it would be tandem of VH and VV. You may calculate VH/VV (VH divided by VV) for each date you have.
Then, if you have NDVI values for the very same dates you would easily obtain a relationship between your VH/VV parameter and NDVI. It has to be something like R2=0.85.
Be aware of the outliers. Some points may decrease your R2 significantly, probably because of the atmospheric issues. May be it wouldn’t be within the limits of scientific approach, but I would recommend to drop off such points.

1 Like