Scaling of Cloud Top Temperature product of MODIS L3

[j.blanco_07]

I am using daily L3 products (for example MYD08_D3.A2003039.061.2018005124705.hdf)

I open the attributes of CTT variable (Cloud_Top_Temperature_Mean.attrs), and what I get is:

{'valid_range': array([ 0, 20000], dtype=int16),
'long_name': 'Cloud Top Temperature: Mean',
'units': 'Degrees Kelvin',
'Level_2_Pixel_Values_Read_As': 'Real',
'Derived_From_Level_2_Data_Set': 'Cloud_Top_Temperature',
'Included_Level_2_Nighttime_Data': 'True',
'Quality_Assurance_Data_Set': 'None',
'Statistic_Type': 'Simple',
'Aggregation_Data_Set': 'None'}

However, neither scale_factor nor add_offset values appear on the list, so how do I compute the actual value in K?

[LAADSx_UserServices_M]

Following information is provided in the HDF file metadata:
MYD08_D3.A2003039.061.2018005124705.hdf
Cloud_Top_Temperature_Mean (720, 288)
16-bit integer, 180 x 360
Number of attributes = 12
valid_range = 0,20000
_FillValue = -9999
long_name = Cloud Top Temperature: Mean
units = Degrees Kelvin
scale_factor = 0.009999999776482582
add_offset = -15000.0
Level_2_Pixel_Values_Read_As = Real
Derived_From_Level_2_Data_Set = Cloud_Top_Temperature
Included_Level_2_Nighttime_Data = True
Quality_Assurance_Data_Set = None
Statistic_Type = Simple
Aggregation_Data_Set = None

[j.blanco_07]

I still have an issue: a global snapshot of the stored CTT ranges between 14960 and 14840.

scale_factor = 0.009999999776482582
add_offset = -15000.0
I use the conversion equation: CTT_actual = scale_factor * ( CTT_stored - add_offset )

now CTT ranges between 0.4 and 1.6 but with units in K this is cannot be right: CTT 's order of magnitude should be ~200 to ~300 K.

I cannot find the error in my calculation.
Looking forward to your reply. Thanks

[LAADSx_UserServices_M]

see:
For stored value = 14840
CTT_actual =0.01*(14840-(-15000)) = 298.4

For stored value = 14960
CTT_actual =0.01*(14960-(-15000)) = 299.6

[j.blanco_07]

Several things:

1) stored CTT ranging between 14960 and 14840 would yield actual CTT ranging 298.4 to 299.6 K. That is less than 2K variability (in a given day) for the entire globe... The example of page 30 in the document "MODISCloudOpticalPropertyUserGuideFinal_v1.1_1.pdf" shows CTT ranging from 220 to 300 K for a subdomain between ~35S and ~50S in the eastern Pacific.
Therefore the scale factor cannot be ~0.01, it should be ~1 (14960 -14840 = 120K)

2) In my previous post I incorrectly reported 14840 and 14960 as the extreme values. Actually I didn't see the minus sign in the colorbar of my plot. Stored CTT ranges between -14960 and -14840. Therefore, both offset and scale factor have to be wrong.

3) I checked files from different years (e.g: MYD08_D3.A2003039.061.2018005124705.hdf, MYD08_D3.A2007365.061.2018045221812.hdf), and different variables (e.g: Cloud_Top_Temperature_Mean, Cloud_Top_Temperature_Day_Mean), but always the same pattern.

4) attributes of stored CTT: "valid_range is [0 20000]", but stored CTT values are negative ??

Really looking forward to the solution of this issue.

-----

By the way: I found this related post for L2 data
viewtopic.php?t=4698

In that case, if stored CTT is 6942, then actual CTT is 219.4K
that stored CTT is in the range: 0, 20000.

I suspect that for L3 product the metadata is wrong, or there is a mismatch between stored data and metadata....

[j.blanco_07]

also checked there is no report regarding this issue at daily L3 data issues webpage: https://atmosphere-imager.gsfc.nasa.gov/issues/daily

[LAADSx_UserServices_M]

There are no negative CTT values for either of the files you mentioned (see attached).
Both offset and scale factor are correct as provided.

[j.blanco_07]

Your help was crucial to find the problem.

The issue was the way I was opening the data: using python, I was treating the files as if they were netcdf, opening them with xarray.open_mfdataset

However, for HDF files, this should not be done, I came to learn. I adapted a script that I found in another forum (https://github.com/pydata/xarray/discussions/6299) and this time the data loaded was just like in your photos above.

Interestingly, I noticed that from both ways of reading the data I was obtaining the same plot , except for changes in the scale (and by setting a different offset value if I used xarray.open_mfdataset, I could get the same plot and same scale)

Thank you !

[LAADSx_UserServices_M]

Glad that you found the solution and the issue is resolved.