Zia model associated to each pixel-processing OLI images

mamontes · by **mamontes** » Tue Feb 19, 2019 2:51 pm America/New_York

Dear SEADAS staff
I need to know what is the flag or variable output I need to check to find out what kind of Zia model is selected for a specific pixel
Thanks a lot in advance for your help!
Martin

OB.DAACx - SeanBailey · by **OB.DAACx - SeanBailey** » Wed Feb 20, 2019 4:52 pm America/New_York

Well, it's a bit complicated now... There are two sets of models which are then averaged (well the averages are averaged). You can see which models are selected and the ratio of the pairs (and ratio of the averages) with the following parameters:

aer_model_1
aer_model_2
aer_model_3
aer_model_4
aer_model_ratio_1_2
aer_model_ratio_3_4
aer_model_ratio

The aer_model_# parameters give you the model number, which is based on it's position in the aermodels list. That list, by default, has 80 models...

Depending upon what you really need (or want), you could use the retrieved angstrom exponent to find the closest model. The l2gen processing log output lists them...or look in the attached file (based on MODIS).

Sean
attachment 1

mamontes · by **mamontes** » Thu Feb 21, 2019 9:12 am America/New_York

Sean
awesome!
thanks a lot and makes a lot sense for me
the alpha AEextinction in that table is computed within which spectral range?
thanks again
Martin

OB.DAACx - SeanBailey · by **OB.DAACx - SeanBailey** » Thu Feb 21, 2019 10:51 am America/New_York

I believe that'd be the 443-869 nm range.

Sean

mamontes · by **mamontes** » Thu Feb 21, 2019 2:26 pm America/New_York

great Sean! thanks

mamontes · by **mamontes** » Thu Feb 21, 2019 2:59 pm America/New_York

sorry Sean, one more. I wonder if this AEextinction440-870 in your attached table was computed with satellite or simulated data?
thanks again
Martin

OB.DAACx - SeanBailey · by **OB.DAACx - SeanBailey** » Thu Feb 21, 2019 3:51 pm America/New_York

It is from the Ahmad-Fraser RT model used in the generation of the aerosol models. So, simulated data.

Sean

mamontes · by **mamontes** » Thu Feb 21, 2019 4:34 pm America/New_York

Sean
great ! and sorry about having one more question
how the aerosol model ratios are computed?? Can I find this info in aerosol.c? ratios of what and which wavelength?
thanks a lot in advance again :)

OB.DAACx - SeanBailey · by **OB.DAACx - SeanBailey** » Thu Feb 21, 2019 7:39 pm America/New_York

During the model selection process, the two aerosol models that bound the measured "epsilon" (ratio of aerosol reflectance in the NIR bands) are selected. The ratio is the distance of the two model epsilon values from the measured one. This is done for two relative humidity sets (the two that bound the RH for the pixel as defined by the input ancillary data). The ratios are essentially the fraction of input from each model in the pair used to derive the aerosol reflectance to be removed from the signal to get to the surface value. Yes, it's somewhere in aerosol.c :)

Sean

mamontes · by **mamontes** » Fri Feb 22, 2019 8:39 am America/New_York

Hi Sean, thanks again. Sorry but still not clear for me.
You have 4 Zia models selected for each pixel corresponding to a specific pixel-derived RH. You have 4 models cause you bracket between 2 RHs, 2 and 2 models for each RH.
Now the confusing part,
1. you compute aerosol reflectance for each Zia model based on LUTs and computed AOD
2.then you compute epsilons for 765 and 865 nm wavelengths (default seadas lambdas traditionally)
3. then you compare these epsilons with the measured epsilon and you keep the 4 having the minimum difference (I guess you select the closest first for each RH and the second closest later) with the measured epsilon for the resulting RH range?
4. then ?? what are these ratios again? the average of epsilons for each RH? why to report them? so basically the FINAL aerosol model is the average of 4 models?

thanks again
Martin