EGU General Assembly 2020
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Classifying Global Low-Cloud Morphology with a Deep Learning Model: Results and Potential Use

Tianle Yuan
Tianle Yuan
  • University of Maryland, JCET, Greenbelt, United States of America (

Marine low clouds display rich mesoscale morphological types, distinct spatial patterns of cloud fields. Being able to differentiate low cloud morphology offers a tool for the research community to go one step beyond bulk cloud statistics such as cloud fraction and advance the understanding of low clouds. Here we report the progress of a NASA funded project that aims to create an observational record of low cloud mesoscale morphology at a near-global (60S-60N) scale. First, a training set is created by our team members manually labeling thousands of mesoscale (128x128) MODIS scenes into six different categories: stratus, closed cellular convection, disorganized convection, open cellular convection, clustered cumulus convection, and suppressed cumulus convection. Then we train a deep convolutional neural network model using this training set to classify individual MODIS scenes at 128x128 resolution, and test it on a test set. The trained model achieves a cross-type average precision of about 93%. We apply the trained model to 16 years of data over the Southeast Pacific. The resulting climatological distribution of low cloud morphology types show both expected and unexpected features and suggest promising potential for low cloud studies as a data product.

How to cite: Yuan, T.: Classifying Global Low-Cloud Morphology with a Deep Learning Model: Results and Potential Use, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-20208,, 2020


Display file

Comments on the display

AC: Author Comment | CC: Community Comment | Report abuse

displays version 1 – uploaded on 01 May 2020
  • CC1: Comment on EGU2020-20208, Tobias Weigel, 06 May 2020

    Very interesting study, the results look promising to me on first glance. How does this relate to the earlier work from Stevens et al. (2019), - are you familiar with this? I'd like to know what exactly you did that improved possibly on those findings (training parameters & data, network architecture, ...).

    • AC1: Reply to CC1, Tianle Yuan, 06 May 2020

      Hi Tobias,

      Thank you for your comments. Yeah, I am aware of the Stevens et al. (2019) work. Their work is very interesting. They are looking at a set of facinating patterns in the tropical trades, specifically. Our work is targeting all low clouds. As a result, we cannot go into specific patterns as defined in their study. We can on the other hand deal with much broader regions. 

      Another difference is the scale. Their method do not seem to be limited by fixed scale. A pattern can cover large areas. Our method deals with a fixed scale, 128x128 pixels. There are pros and cons for both approaches. 

      In short, I think two methods are quite complementary to each other. It would be interesting to compare and contrast results!