14 Nov 2022
14 Nov 2022
Status: this preprint is open for discussion.

CLGAN: A GAN-based video prediction model for precipitation nowcasting

Yan Ji1,2, Bing Gong2, Michael Langguth2, Amirpasha Mozaffari2, and Xiefei Zhi1 Yan Ji et al.
  • 1Nanjing University of Information Science and Technology, 210044 Nanjing, China
  • 2Jülich Supercomputing Centre, Forschungszentrum Jülich, 52425 Jülich, Germany

Abstract. The prediction of precipitation patterns at high spatio-temporal resolution up to two hours ahead, also known as precipitation nowcasting, is of great relevance in weather-dependant decision-making and early warning systems. In this study, we are aiming to provide an efficient and easy-to-understand model - CLGAN, to improve the nowcasting skills of heavy precipitation events with deep neural networks for video prediction. The model constitutes a Generative Adversarial Network (GAN) architecture whose generator is built upon an u-shaped encoder-decoder network (U-Net) equipped with recurrent LSTM cells to capture spatio-temporal features. A comprehensive comparison among CLGAN, and baseline models optical flow model DenseRotation as well as the advanced video prediction model PredRNN-v2 is performed. We show that CLGAN outperforms in terms of scores for dichotomous events and object-based diagnostics. The ablation study indicates that the GAN-based architecture helps to capture heavy precipitation events. The results encourage future work based on the proposed CLGAN architecture to improve the precipitation nowcasting and early-warning systems.

Yan Ji et al.

Status: open (until 09 Jan 2023)

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

Yan Ji et al.

Yan Ji et al.


Total article views: 215 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
160 49 6 215 3 2
  • HTML: 160
  • PDF: 49
  • XML: 6
  • Total: 215
  • BibTeX: 3
  • EndNote: 2
Views and downloads (calculated since 14 Nov 2022)
Cumulative views and downloads (calculated since 14 Nov 2022)

Viewed (geographical distribution)

Total article views: 214 (including HTML, PDF, and XML) Thereof 214 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
Latest update: 06 Dec 2022
Short summary
Formulating the short-term precipitation forecasting as a video prediction task, a novel deep learning architecture CLGAN is proposed in this work. A benchmark data set is newly built for the task, on minute-level precipitation measurements. Our results show that the GAN-component of CLGAN encourages the model to generate predictions sharing statistical properties of observed precipitation, which makes it outperform the baseline in dichotomous and spatial scores for heavy precipitation events.