| 23 Oct 2008 10:40 AM |
|
|
I am a student from China, now I have some problems in dealing with FLAASH.
I want to correct a MODIS image by FLAASH model,
The data is produced by "MODIS Converse Toolkit", I think the input data is no problem,
It is always get an error.
"ACC error: lsmooth2:
IDL error:
ACC_LSMOOTH2: Cannot continue with smoothing calculation
"
But I can setting "use adjacency correction" NO , this error will be fine. What should I do?
And I have some other questions.
The scene center location means what? Is the imaging center of the MODIS data.
If I subset the 2030*1354 data to 100*200, the scene center location is the position of the
(1015,677) point or the new center point of the new 100*200 data?
In the "advanced settings" the non-nadir zenith angle and azimuth angle means what?
The user guide said the zenith angle is between 90 and 180, but I find in the modis data
I used the zenith angle stored in the MOD02 "zenith angle" dataset, the center location's
Zenith angle is only 4. should I use the value of 180 minus 4?
When I input the input radiance image, it will prompt me to set a scale factor,
10 or 0.1 which value I should to set, I was puzzled.
The last one is the "water retrieval" in the multispectral settings there is a "index to first band",
I don't know this means what. And I choose the water absorption feature 940 nm, in the multispectral settings
I set the band19 and band 16 for absorption and reference channel respectively, and it also always count an error.
"ACC error: h2oinv: h2oinv
IDL error
Array has too many elements
"
|
|
|
|
MariM
Veteran Member
Posts:2396  
| 28 Oct 2008 09:23 AM |
|
|
It is fine to set the adjacency correction to No for MODIS data. You cannot calculate an adjacency correction for large pixels because the scattered light stays within the original pixel.
The scene center is the approximate center of the image. I would use the original center from a subset image.
The non-nadir viewing can be set for sensors that are not pointing at nadir (straight down). You can add this setting to improve accuracy for off-nadir imagery. For MODIS, this is generally not a setting that is used. If you do set this, 180 is considered nadir, so you would use 176 for a 4 degree off-nadir view.
The scale factor is the value that will be divided into the data to convert it into the expected units of microW/(cm^2/sr/nm). So you need to know the radiance units the data is in to know the scale factor to use.
You may or may not be able to achieve a water retrieval with MODIS data. If you include all 19 bands, it may be possible but it depends on whether the water absorption feature is pronounced enough to be detected. You may need to try the correction without the water retrieval. The 'index to first band' is used to make sure the bands in the input file match those in the MODIS filter function. It is important that they match band for band in order to achieve a good correction.
|
|
|
|
Deleted User
New Member
Posts: 
| 29 Oct 2008 08:13 AM |
|
|
Thanks for mminari's reply.
you say "I would use the original center from a subset image." means choose the original center as the center for the subseted one? for example MOD1KM is 2030*1354 and I subset a 200*200 from it ,the center point of the 200*200 image I should choose the (1015,677)? or the (100,100) of the 200*200 image?
about the scale factor , the unit of the calibrated MODIS data is W/(m^2/sr/um), while the FLAASH required is
microW/(cm^2/sr/nm), the unit of MODIS is 1/10 of FLAASH required, I think it will be input 0.1 while some files I find say it should input 10.
and I think there is an little error in the techtip of 4093, the band 7 of MODIS is about 2.1um
the techtip of 4093 say" Aerosol retrieval cannot be performed because it requires a band in the 2.1 micrometer range."
|
|
|
|
MariM
Veteran Member
Posts:2396
| 29 Oct 2008 10:01 AM |
|
|
Typically I would use the lat/lon of approximately the center of the original image, not the subset. However, MODIS covers a large area so you could also use the approximate center of the subset as well. The point is just to allow FLAASH to know the approximate latitude of the input data.
A simple conversion from W/m^2*micrometer to microW/cm^2*nm is a scale factor of 10.
Performing an aerosol retrieval on multispectral data is similar to performing a water retrieval. You can try an aerosol retrieval but there is only a single band near 2.1 micrometers.
|
|
|
|
Deleted User
New Member
Posts:
| 29 Oct 2008 11:47 PM |
|
|
I have another question.
Do you have any idea on how to check the FLAASH model is rightly processed.
the image of processed before and after is very similar.
You say "You may or may not be able to achieve a water retrieval with MODIS data.
If you include all 19 bands, it may be possible but it depends on whether the water absorption
feature is pronounced enough to be detected. "
what does this mean, water retrieval succussfully or not is depends on the MODIS data itself?
|
|
|
|
MariM
Veteran Member
Posts:2396
| 30 Oct 2008 07:43 AM |
|
|
Yes, FLAASH performs water retrieval by estimating the water absorption across certain portions of the spectrum. If the water absorption feature cannot be detected because there is not sufficient spectral resolution to detect it, then a pixel-by-pixel correction for water cannot be obtained. This is why it works best for hyperspectral data in which there is sufficient spectral resolution to make these estimates.
|
|
|
|
Deleted User
New Member
Posts:
| 01 Nov 2008 03:38 AM |
|
|
Under mminari's instruction, I have corrected the MODIS data by FLAASH without error.
And at the same time I retrieve the water vapor using the band 19 and band 2 as the absorption and reference band
respectively.
but I am NOT sure the result I get is right or not. on the whole the reflectance value of the corrected data is more than 20000 times of the value of original data.
the most important point is the reflectance of water body is also more than 20000 times of that of the original data.
For example I used a MODIS data cover Bo sea of CHINA,
a pixel in the sea the reflectance value of band 7,2,1 are 43, 194, 432(multiple by 10000)
and the value after the FLAASH model of these bands are 81,333,824(output data is also multiple by 10000 setting in Advance setting of FLAASH) .
why this happened, the value after atmospheric correction should smaller than before.
And the water retrieved by FLAASH have some bad value -224, what this value's mean,
on the whole the value retrieved by FLAASH have a acceptable value compared with MOD05 water vapor products
HOW could I validate the result of the FLAASH model.
Thank you in advance.
|
|
|
|