Preprints
https://doi.org/10.5194/egusphere-2024-1336
https://doi.org/10.5194/egusphere-2024-1336
22 May 2024
 | 22 May 2024
Status: this preprint is open for discussion and under review for Atmospheric Measurement Techniques (AMT).

Severe hail detection with C-band dual-polarisation radars using convolutional neural networks

Vincent Forcadell, Clotilde Augros, Olivier Caumont, Kévin Dedieu, Maxandre Ouradou, Cloe David, Jordi Figueras i Ventura, Olivier Laurantin, and Hassan Al-Sakka

Abstract. Radar has consistently proven to be the most reliable source of information for the remote detection of hail within storms in real-time. Currently, existing hail detection techniques have limited ability to clearly distinguish storms that produce severe hail from those that do not. This often results in a prohibitive number of false alarms that hamper real-time decision-making. This study utilises convolutional neural network (CNN) models trained on dual-polarisation radar data to detect severe hail occurrence on the ground. The morphology of the storms is studied by leveraging the capabilities of a CNN. A database of images of 60 km x 60 km containing 19 different radar-derived features is built above severe hail reports (above 2 cm) and above rain or small hail reports (rain or hail below 2 cm) created for the occasion with the help of a cell-identification algorithm. After a tuning phase on the CNN architecture and its input size, the CNN is trained to output one probability of severe hail on the ground per image of 30 km x 30 km. A test set of 1396 images between 2018 and 2023 demonstrates that the CNN method outperforms state-of-the-art methods according to various metrics. A feature importance study indicates that existing hail proxies as input features are beneficial to the CNN, particularly the maximum estimated size of hail (MESH). The study demonstrates that many of the existing radar hail proxies can be adjusted using a threshold value and a threshold area to achieve similar performance to that of the CNN for severe hail detection. Finally, the output of ten fitted CNN models in inference mode on a hail event is shown.

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.
Vincent Forcadell, Clotilde Augros, Olivier Caumont, Kévin Dedieu, Maxandre Ouradou, Cloe David, Jordi Figueras i Ventura, Olivier Laurantin, and Hassan Al-Sakka

Status: open (until 27 Jun 2024)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
Vincent Forcadell, Clotilde Augros, Olivier Caumont, Kévin Dedieu, Maxandre Ouradou, Cloe David, Jordi Figueras i Ventura, Olivier Laurantin, and Hassan Al-Sakka
Vincent Forcadell, Clotilde Augros, Olivier Caumont, Kévin Dedieu, Maxandre Ouradou, Cloe David, Jordi Figueras i Ventura, Olivier Laurantin, and Hassan Al-Sakka

Viewed

Total article views: 172 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
123 42 7 172 7 8
  • HTML: 123
  • PDF: 42
  • XML: 7
  • Total: 172
  • BibTeX: 7
  • EndNote: 8
Views and downloads (calculated since 22 May 2024)
Cumulative views and downloads (calculated since 22 May 2024)

Viewed (geographical distribution)

Total article views: 166 (including HTML, PDF, and XML) Thereof 166 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 
Latest update: 12 Jun 2024
Download
Short summary
This study demonstrates the potential for enhancing severe hail detection through the application of convolutional neural networks (CNNs) to dual-polarization radar data. It is shown that current methods can be calibrated to significantly enhance their performance for severe hail detection. This study establishes the foundation for the solution of a more complex problem: the estimation of the maximum size of hailstones on the ground using deep learning applied to radar data.