Unsupervised domain adaptation for Global Precipitation Measurement satellite constellation using Cycle Generative Adversarial Nets - Archive ouverte HAL Access content directly
Journal Articles Environmental Data Science Year : 2022

Unsupervised domain adaptation for Global Precipitation Measurement satellite constellation using Cycle Generative Adversarial Nets

(1) , (1) , (1) , (1) , (1)
1
Vibolroth Sambath
Nicolas Viltard
Connectez-vous pour contacter l'auteur
Laurent Barthès
  • Function : Author
  • PersonId : 968696
Audrey Martini
  • Function : Author
  • PersonId : 946557
Cécile Mallet

Abstract

Artificial intelligence has provided many breakthroughs in the field of computer vision. The fully convolutional networks U-Net in particular have provided very promising results in the problem of retrieving rain rates from spaceborne observations, a challenge that has persisted over the past few decades. The rain intensity is estimated from the measurement of the brightness temperatures on different microwave channels. However, these channels are slightly different depending on the satellite. In the case where a retrieval model has been developed from a single satellite, it may be advantageous to use domain adaptation methods in order to make this model compatible with all the satellites of the constellation. In this proposed feasibility study, a Cycle Generative Adversarial Nets model is used for adapting one set of brightness temperature channels to another set. Results of a toy experiment show that this method is able to provide qualitatively good precipitation structure but still could be improved in terms of precision. Impact Statement Supervised deep learning approaches in climate studies, especially in satellite observations, are very limited in application due to the non-conventional nature of the data and the lack of available annotated samples. The present feasibility study on unsupervised domain adaptation aims to increase the compatibility of a deep learning model pre-trained on one satellite to many more with similar physical characteristics. While previous approaches focus on qualitative aspects and classification tasks, the present objective involves a regression task on non-RGB (red, green, and blue) image data. The adaptation results significantly impact the practical perspectives of applying deep learning models to the spatial observation of the earth. In terms of climate studies, this unsupervised transfer learning approach will improve the knowledge of the precipitation evolution over the last 30 years.
Fichier principal
Vignette du fichier
unsupervised-domain-adaptation-for-global-precipitation-measurement-satellite-constellation-using-cycle-generative-adversarial-nets.pdf (6.87 Mo) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

insu-03892915 , version 1 (10-12-2022)

Licence

Attribution - NonCommercial - CC BY 4.0

Identifiers

Cite

Vibolroth Sambath, Nicolas Viltard, Laurent Barthès, Audrey Martini, Cécile Mallet. Unsupervised domain adaptation for Global Precipitation Measurement satellite constellation using Cycle Generative Adversarial Nets. Environmental Data Science, 2022, 1, pp.e24. ⟨10.1017/eds.2022.16⟩. ⟨insu-03892915⟩
0 View
0 Download

Altmetric

Share

Gmail Facebook Twitter LinkedIn More