Earthdata Forum

Hi all,

I am wondering to filter low-level (cloud top pressure >= 680 hPa) and warm (cloud top temperature > 273.0 K) cloud fraction in MCD06COSP level-3 daily cloud properties (https://ladsweb.modaps.eosdis.nasa.gov/missions-and-measurements/products/MCD06COSP_D3_MODIS).

I can see two different flavours of cloud fraction ('Cloud Mask Fraction Low' and 'Cloud Retrieval Fraction Liquid') but not sure about selecting only low-level warm cloud fraction.
Since MCD06COSP dataset also provide 'cloud top pressure' as one of SDS, may I choose 'Cloud Retrieval Fraction Liquid' where 'cloud top pressure' >= 680 hPa. Any suggesting from experts!
Thanks

Below is a response from our SME:

Cloud Mask Fraction Low: it is based on cloud top pressure (CTP >= 680hpa) for daytime (SZA <= 85 deg). The daytime cutoff is less strict than cloud optical properties (SZA <= 81.3731 deg), so the data extends a bit farther poleward. It may include liquid, ice, or undetermined phase cloud.

Cloud Retrieval Fraction Liquid: is based on cloud phase for liquid cloud only and daytime (SZA <= 81.3731 deg). The daytime cutoff is slightly more restrictive than Cloud Mask Fraction.

Therefore, for low-level warm cloud fraction, we suggest using the overlapped pixels of “Cloud Mask Fraction Low” AND “Cloud Retrieval Fraction Liquid”.

Okay, I will consider overlapping pixels of “Cloud Mask Fraction Low” AND “Cloud Retrieval Fraction Liquid” for low-level warm cloud selection.
Thanks for your kind suggestion.

I am using python (xarray and numpy)
mask = 1 * numpy.ones((180,360)) * numpy.isfinite(Cloud_Mask_Fraction_Low)
low_warm_cloud_fraction = Cloud_Retrieval_Fraction_Liquid.where(mask)

is that correct way?

Yes, this is a correct way.
The logic is to find the pixels with valid cloud fraction values for both Cloud_Mask_Fraction_Low and Cloud_Retrieval_Fraction_Liquid, then for these pixels you can use either Cloud_Retrieval_Fraction_Liquid or Cloud_Mask_Fraction_Low.

Thanks for your kind suggestion!

one more thing I want to ask, while computing cloud fraction in MODIS COSP dataset, do the computation consider the effect of overlap cloud layer? Some studies (e.g., DOI: 10.1126/sciadv.adh771) while using MODIS cloud fraction prefer random overlap assumption as:
CF_low = (CF_total - CF_high) / (1 - CF_high)
where CF_low, CF_high and CF_total are low cloud, high cloud and total cloud fraction.
Any comment on it!
Thanks

In MODIS L3 COSP, the cloud fractions are aggregated at different pressure levels, there is no layer overlap:
Cloud_Mask_Fraction: (all pressure levels: total column)
Cloud_Mask_Fraction_Low: (CTP >= 680 hpa)
Cloud_Mask_Fraction_Mid: (440 hpa <= CTP < 680 hpa)
Cloud_Mask_Fraction_High: (CTP < 440 hpa)
Cloud_Mask_Fraction = Cloud_Mask_Fraction_Low + Cloud_Mask_Fraction_Mid + Cloud_Mask_Fraction_High

For L3 cloud fraction derived from cloud optical properties:
Cloud_Retrieval_Fraction_Total = Cloud_Fraction_Liquid + Cloud_Fraction_Ice + Uncetermined_Phase_Cloud

Thanks for the clarification.