04 Aug 2022
04 Aug 2022
Status: this preprint is open for discussion.

Deep learning models for generation of precipitation maps based on Numerical Weather Prediction

Adrian Rojas-Campos1, Michael Langguth2, Martin Wittenbrink3, and Gordon Pipa1 Adrian Rojas-Campos et al.
  • 1Institute of Cognitive Science, Osnabrück University, 49069 Osnabrück, Germany
  • 2Juelich Supercomputing Centre, Forschungszentrum Juelich, 52425 Jülich, Germany
  • 3Deutscher Wetterdienst, 63067 Offenbach, Germany

Abstract. Numerical Weather Prediction models (NWP) are atmospheric simulations that imitate the dynamics of the atmosphere and provide high-quality forecasts. One of the most significant limitations of NWP is the elevated amount of computational resources required for its functioning, which limits the spatial and temporal resolution of the outputs. Traditional meteorological techniques to increase the resolution are based uniquely on information from a limited group of interest variables. In this study, we offer an alternative approach to the task where we generate precipitation maps based on the complete set of variables of the NWP to generate high-resolution and short-time precipitation predictions. To achieve this, five different deep learning models were trained and evaluated: baseline, U-Net, two deconvolutional networks, and one conditional generative model (CGAN). A total of 20 independent random initializations were performed for each of the models. The predictions were evaluated using MAE and LEPS-based skill scores, ETS, CSI, and frequency bias after applying several thresholds. The models showed a significant improvement in predicting precipitation showing the benefits of including the complete information from the NWP. The algorithms increased the resolution of the predictions and corrected an over-forecast bias from the input information. However, some new models presented new types of bias: U-Net tended to mid-range precipitation events, and the deconvolutional models favored low rain events and generated some spatial smoothing. The CGAN offered the highest quality precipitation forecast generating realistic outputs and indicating possible future research paths.

Adrian Rojas-Campos et al.

Status: open (extended)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
  • CEC1: 'Comment on egusphere-2022-648', Juan Antonio Añel, 24 Aug 2022 reply
    • AC1: 'Reply on CEC1', Adrian Rojas Campos, 04 Sep 2022 reply
      • CEC2: 'Reply on AC1', Juan Antonio Añel, 13 Oct 2022 reply
        • AC4: 'Reply on CEC2', Adrian Rojas Campos, 24 Oct 2022 reply
  • RC1: 'Comment on egusphere-2022-648', Anonymous Referee #1, 05 Sep 2022 reply
    • AC2: 'Reply on RC1', Adrian Rojas Campos, 27 Sep 2022 reply
  • EC1: 'Comment on egusphere-2022-648', Chanh Kieu, 08 Oct 2022 reply
    • AC3: 'Reply on EC1', Adrian Rojas Campos, 12 Oct 2022 reply

Adrian Rojas-Campos et al.

Adrian Rojas-Campos et al.


Total article views: 726 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
530 174 22 726 4 4
  • HTML: 530
  • PDF: 174
  • XML: 22
  • Total: 726
  • BibTeX: 4
  • EndNote: 4
Views and downloads (calculated since 04 Aug 2022)
Cumulative views and downloads (calculated since 04 Aug 2022)

Viewed (geographical distribution)

Total article views: 688 (including HTML, PDF, and XML) Thereof 688 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
Latest update: 29 Nov 2022
Short summary
Our manuscript presents an alternative approach for generating high-resolution precipitation maps based on the non-linear combination of the complete set of variables of the numerical weather predictions. This process combines the super-resolution task with the bias correction in a single step, generating high-resolution corrected precipitation maps with 3 hour lead time. We used using deep learning algorithms to combine the input information and increase the accuracy of the precipitation maps.