High albedo problems during OpticalCalibration

#1

I have a WorldView 2 scene from an arid region of South Africa where I want to map Quartz fields. These are areas where quartz stones are forming large patches with mostly without “visible” vegetation. I used the otbcli_OpticalCalibration Tool to transform the DNs to reflectances per pixel (=TOA). my code:

otbcli_OpticalCalibration ^ -in c:\Users\BAF1858\Downloads\059176997010_01\MUL\17APR11091403-M2AS-59176997010_01_P001.TIF ^ -out c:\Users\BAF1858\Downloads\059176997010_01\MUL\17APR11091403-M2AS-TOA.TIF ^ -level toa ^ -atmo.aerosol desertic

I could have also entered all the information from the metadata-file, but the result is the same.

The problem is now that there are many “blue spots” in the TOA-image (the lower image), but only where the albedo is high. Does anybody have an idea why this is so, and how one could get rid of these errors? I was thinking of smoothing or so, but maybe someone knows a fix for the calibration tool?

While checking the images more closely, I also found that the NDVI throws out values between -0.61 and 0.17. Also the spectral profile for a clearly green pixel behaves strange. See below:

0 Likes

#2

I found the error - adding the argument -atmo.aerosol desertic seems to lead to the errors. Without this argument, everything looks good. This is strange, because this area is clearly an arid region (though not a classical desert). Anyway - problem solved. :clap:

0 Likes

#3

I think that the aerosol model is only used in toc (Top Of Canopy) level, and you are using toa (Top of Atmosphere) level.

1 Like