Brief communication: Training of AI-based nowcasting models for rainfall early warning should take into account user requirements

Ayzel, Georgy; Heistermann, Maik

doi:https://doi.org/10.5194/nhess-25-41-2025

Articles | Volume 25, issue 1

https://doi.org/10.5194/nhess-25-41-2025

© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/nhess-25-41-2025

© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 25, issue 1

Brief communication

|

03 Jan 2025

Brief communication |

| 03 Jan 2025

Brief communication: Training of AI-based nowcasting models for rainfall early warning should take into account user requirements

Georgy Ayzel and Maik Heistermann

Supplement

https://doi.org/10.5194/nhess-25-41-2025-supplement

Data sets

Reprocessed quasi gauge-adjusted radar data, 5-minute precipitation sums (YW) T. Winterrath et al. https://doi.org/10.5676/DWD/RADKLIM_YW_V2017.002

Heavy precipitation events Version 2021.01 exceeding DWD warning level 3 for severe weather based on RADKLIM-RW Version 2017.002 K. Lengfeld et al. https://doi.org/10.5676/DWD/CatRaRE_W3_Eta_v2021.01

Model code and software

The RainNet2024 family of models for precipitation nowcasting Georgy Ayzel https://doi.org/10.5281/zenodo.12547127

Download

Article (1347 KB)
Full-text XML

Short summary

Forecasting rainfall over the next hour is an essential feature of early warning systems. Deep learning (DL) has emerged as a powerful alternative to conventional nowcasting technologies, but it still struggles to adequately predict impact-relevant heavy rainfall. We think that DL could do much better if the training tasks were defined more specifically and that such specification presents an opportunity to better align the output of nowcasting models with actual user requirements.