07 Feb 2024
 | 07 Feb 2024
Status: this preprint is open for discussion and under review for Atmospheric Measurement Techniques (AMT).

Marine cloud base height retrieval from MODIS cloud properties using machine learning

Julien Lenhardt, Johannes Quaas, and Dino Sejdinovic

Abstract. Clouds are a crucial regulator in the Earth's energy budget through their radiative properties, both at the top-of-the-atmosphere and at the surface, hence determining key factors like their vertical extent is of essential interest. While the cloud top height is commonly retrieved by satellites, the cloud base height is difficult to estimate from satellite remote sensing data. Here we present a novel method leveraging spatially resolved cloud properties from the MODIS instrument to retrieve the cloud base height over marine areas. A machine learning model is built with two components to facilitate the cloud base height retrieval: the first component is an autoencoder designed to learn a representation of the data cubes of cloud properties and reduce their dimensionality. The second component is developed for predicting the cloud base using ground-based ceilometer observations from the lower dimensional encodings generated by the aforementioned autoencoder. The method is then evaluated based on a collection of co-located surface ceilometer observations and retrievals from the CALIOP satellite lidar. The statistical model performs well on both datasets, exhibiting accurate predictions in particular for lower cloud bases and a narrow distribution of the absolute error, namely 379 m and 328 m for the mean absolute error and the standard deviation of the absolute error respectively for cloud bases in the test set. Furthermore, cloud base height predictions are generated for an entire year over ocean, and global mean aggregates are also presented, providing insights about global cloud base height distribution and offering a valuable dataset for extensive studies requiring global cloud base height retrievals. The global cloud base height dataset and the presented models are available from Zenodo (Lenhardt et al., 2024).

Julien Lenhardt, Johannes Quaas, and Dino Sejdinovic

Status: open (until 06 Apr 2024)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
Julien Lenhardt, Johannes Quaas, and Dino Sejdinovic
Julien Lenhardt, Johannes Quaas, and Dino Sejdinovic


Total article views: 186 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
141 41 4 186 1 1
  • HTML: 141
  • PDF: 41
  • XML: 4
  • Total: 186
  • BibTeX: 1
  • EndNote: 1
Views and downloads (calculated since 07 Feb 2024)
Cumulative views and downloads (calculated since 07 Feb 2024)

Viewed (geographical distribution)

Total article views: 181 (including HTML, PDF, and XML) Thereof 181 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
Latest update: 04 Mar 2024
Short summary
Clouds play a key role in the regulation of the Earth's climate. Aspects like the height of their base are of essential interest, but remain difficult to derive from satellite data. In this study, we combine observations from the surface and satellite retrievals of cloud properties to build a robust and accurate method to retrieve the cloud base height.