04 Jul 2022
04 Jul 2022
Status: this preprint is currently under review for the journal AMT.

Ice water path retrievals from Meteosat-9 using quantile regression neural networks

Adrià Amell, Patrick Eriksson, and Simon Pfreundschuh Adrià Amell et al.
  • Department of Space, Earth and Environment, Chalmers University of Technology, Gothenburg, Sweden

Abstract. The relationship between geostationary radiances and ice water path (IWP) is complex, and traditional retrieval approaches are not optimal. This work applies machine learning to improve the IWP retrieval from Meteosat-9 observations, with a focus on low latitudes, training the models against retrievals based on CloudSat. Advantages of machine learning include avoiding explicit physical assumptions on the data, an efficient use of information from all channels, and easily leveraging spatial information.

Thermal infrared (IR) retrievals are used as input to achieve a performance independent of the solar angle. They are compared with retrievals including solar reflectances, as well as a subset of IR channels for compatibility with historical sensors. The retrievals are accomplished with quantile regression neural networks. This network type provides case-specific uncertainty estimates, compatible with non-Gaussian errors, and is flexible enough to be applied to different network architectures.

Spatial information is incorporated into the network through a convolutional neural network (CNN) architecture. This choice outperforms architectures that only work pixelwise. In fact, the CNN shows a good retrieval performance by using only IR channels. This allows computing diurnal cycles, a problem that CloudSat cannot resolve due to its limited temporal and spatial sampling. These retrievals compare favourably with IWP retrievals in CLAAS, a thoroughly validated dataset based on a traditional approach. These results highlight the possibilities to overcome limitations from physics-based approaches using machine learning while providing efficient, probabilistic IWP retrieval methods. Moreover, they suggest extending this first work to higher latitudes as well as considering geostationary data as a complement to the upcoming Ice Cloud Imager mission, for example, to bridge the gap in temporal sampling with respect to space-based radars.

Adrià Amell et al.

Status: open (until 20 Aug 2022)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse

Adrià Amell et al.

Data sets

Ice water path retrievals from Meteosat-9 with quantile regression neural networks: code and models Adrià Amell

Video supplement

Ice water path retrievals from Meteosat-9 with quantile regression neural networks: video supplement Adrià Amell

Adrià Amell et al.


Total article views: 221 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
170 45 6 221 2 2
  • HTML: 170
  • PDF: 45
  • XML: 6
  • Total: 221
  • BibTeX: 2
  • EndNote: 2
Views and downloads (calculated since 04 Jul 2022)
Cumulative views and downloads (calculated since 04 Jul 2022)

Viewed (geographical distribution)

Total article views: 207 (including HTML, PDF, and XML) Thereof 207 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
Latest update: 08 Aug 2022
Short summary
Geostationary satellites continuously image a given location on Earth, a feature that satellites designed to characterize atmospheric ice lack. However, the relationship between geostationary images and atmospheric ice is complex. Machine learning is used here to leverage such images to characterize atmospheric ice throughout the day in a probabilistic manner. Using structural information from the image improves the characterization, and this approach compares favourably to traditional methods.