Earthdata Forum

I am exploring spatial variability of AOD at the centre of a granule/swath of MODIS 04 Level 2 data, particularly Aqua data so that I can compare it with the data extracted from CALIOP. But, I can see that most of the AOD values along the centre (column) are missing. Is there any specific reason for this? Is there a way to address the missing data ?
Also, the coverage of the MODIS swath keeps on changing everyday. How can we address this issue if we need AOD from a particular location over a time period, let's say a month or year?

Many thanks in advance,
Laxmi

Dear Laxmi:
Can you provide an example from files that exhibit such behavior (couple of file names should be sufficient)?

Thank you so much for your reply! Some of the file names include: MYD04_L2.A2021001.1840.061.2021002163532.hdf, MYD04_L2.A2021001.2015.061.2021002164636.hdf, MYD04_L2.A2021002.1920.061.2021003155147.hdf, MYD04_L2.A2021003.2005.061.2021004170411.hdf. I am attaching a screenshot for your reference. Many thanks in advance!

Dear Laxmi:
Please note the following:
Have you tried to cross-reference the missing areas with the MODIS cloud mask (MxD35_L2) to see if those areas have persistent cloud. You can also reference a RGB reflectance image for this.

With regard to this question:
coverage of the MODIS swath keeps on changing everyday. How can we address this issue if we need AOD from a particular location over a time period, let's say a month or year?

Answer: MODIS follows a 16-day repeat cycle, so they can revisit the same area, every 16-days to get AOD observations over the same at-nadir area. But also let them know that following the last set of Inclination maneuvers conducted for Terra (in early 2020) and for Aqua (Spring of 2021), their orbits are slowly drifting, affecting the exact repeatability of observations of the same area. You can find more information in this regard here:
https://landweb.modaps.eosdis.nasa.gov/cgi-bin/QA_WWW/displayCase.cgi?esdt=MOD&caseNum=PM_MOD_21264&caseLocation=cases_data&type=C6

and here:
https://landweb.modaps.eosdis.nasa.gov/cgi-bin/QA_WWW/displayCase.cgi?esdt=MYD&caseNum=AS_MYD_21278&caseLocation=cases_data&type=C6

Also, the Dark Target AOD cannot be retrieved over brighter surfaces. So, you may be getting confused by an apparent pattern that forms over the equatorial and tropical areas, where most of the areas are brighter because of clouds and deserts, and so there is not much retrieval of DT AOD, creating often empty patches across the middle of the swath. The sun-glint over those equatorial areas is another big factor.

We further looked into the granules to check the Aerosol Optical Depth (AOD) pattern you reported and below is the summery:

The Missing Aerosol Optical Data (AOD) is not the systematic issue. ​
They are relevant to the pixel selections. For the following conditions, the AOD values will be​ fill values:​

Over Ocean:​
· Bright Ocean reflectance; ​
· High cloud fraction​
· Over ocean: glint angle < 40 deg and Refl_0.47/Refl_0.66 >= 0.95​
· If AOD > 5 or AOD < -0.01, then report fill values.​

Over Land:​
· If angles and reflectance values are invalid, then AOD will be value.​
· For cloudy and snow/ice pixels.​
· For non “dark target pixels” (Refl_2.13um is outside [0.01, 0.25]​
· According to ρ0.66, discard brightest 50% and darkest 25% of pixels, ​
· If AOD > 5 or AOD < -0.01, then report fill values.

Dear LAADS_UserServices_M,

Thank you so much for the detailed explanation!
Basically, swaths in the Pacific regions are missing AOD values around the middle column of the granules. In terms of fill values, just to be clear with the concept, one of the granules has following details:
{'Cell_Across_Swath_Sampling': [1, 1354, 10],
'Cell_Along_Swath_Sampling': [1, 2021, 10],
'Geolocation_Pointer': 'Internal geolocation arrays',
'Parameter_Type': 'Output',
'_FillValue': -9999,
'add_offset': 0.0,
'long_name': 'Combined Dark Target, Deep Blue AOT at 0.55 micron for land and ocean.',
'scale_factor': 0.0010000000474974513,
'units': 'None',
'valid_range': [-100, 5000]}

Isn't this fill value already is added during the AOD values retrieval for the condition you have mentioned previously? And, in this case -9999 is basically the one to be removed from the data?

Many thanks in advance for your response!

Kind Regards,
Laxmi

Yes, you are correct.

Dear LAADS_UserServices_Mm

Many thanks for the quick response and explanation. In such missing case, would it be a good idea to fill up the missing values by making sub-grid and applying different imputation techniques?
As an additional query, I am wondering if there is an example python code or case studies extracting time series AOD Values for multiple location all at once?

Kind Regards,
Laxmi

No, there are no python code examples available at LAADS. However, you may want to check with "HDF-EOS TOOLS AND INFORMATION CENTER", they have Python code example to work with MODIS data. Their website is
https://www.hdfeos.org/