Karen Kowalewsky
New Member
Posts:9  
| 25 Nov 2022 11:53 AM |
|
|
I am working with some sensor test data and need to atmospherically correct it. When I run QUAC, no problems with the data converting from radiance to reflectance. However, when I run FLAASH (ENVI 5.6.3 and ENVI 5.6.2) I get negative values in the blue to green region of the spectrum. The radiance data is in W/cm-2 - sr - um and I use the conversion/gain factor of 1000 to convert to um/cm-2 - sr - nm for FLAASH. The data is float point data in BIP format. Conditions are MLS profile, urban aerosols, using 1135 nm band for water retrieval, DISORT 4 stream and MODTRAN resolution of 1 cm-1.
To test to see if I was doing something wrong, I ran the AVIRIS 2011 tutorial, followed the instructions to the word and also received very low reflectance values. What am I doing wrong?
I have example images to show what is happening - How do I post those?
|
|
|
|
MariM
Veteran Member
Posts:2396 
| 28 Nov 2022 06:51 AM |
|
|
Unfortunately, our website forum will not allow image uploads and posting.
It is very common to have negative reflectance values returned in FLAASH, particularly in pixels with very low radiance such as shadow and water.
What are the radiance values in these regions? Are they very low? Typical radiance values in microW/cm^2 will range from 1-30 or so.
Quac will automatically set any value <0 to be 0 so you do not have negative values while FLAASH will not so that you can see where pixels go negative. I often suggest to users that they mask any FLAASH pixels that fall below 0 as it is done in Quac.
Also do remember that the values are scaled by 10000, so differences in surface reflectance are often very small when converted to floating point reflectance between 0-100%.
If you believe your image to have low haze and aerosols (very clear), you can change the initial visibility value to something like 100km to 'boost' the blue and green bands.
|
|
|
|
Karen Kowalewsky
New Member
Posts:9
| 28 Nov 2022 09:41 AM |
|
|
Hi Mari - Thanks for the response. I am aware of the shadow and water issue and know that aerosol visibility has a role, but this is for vegetation pixels with no shadow and this odd spectrum is happening for all pixels in the scene - vegetation, soils, man-made materials, etc. I'll run some statistics on the radiance to look for negative or zero values. This data/image is synthetic imagery created using MODTRAN with mlw profile, urban and 15 km visibility. I ran FLAASH for the data developed using mlw profile, urban and 30 km visibility and had the same results in the blue-green part of the spectrum, negative.
|
|
|
|
MariM
Veteran Member
Posts:2396
| 28 Nov 2022 10:12 AM |
|
|
You might post the stats of the input image but if it is only some pixels and most the majority that are negative, it may be typical. If there are negative pixels *in* a healthy green vegetation in say, NIR, then I would think the calibration is off. Remember, the scale factor you use in FLAASH is *divided* into the data cube. So if you look at your data values in radiance and then divide by your scale factor, the resulting radiance values should range from around 1-30 or so.
|
|
|
|