Detail publikačního výsledku

Fully differentiable Lagrangian convolutional neural network for physics-informed precipitation nowcasting

PAVLÍK, P.; VÝBOH, M.; BOU EZZEDDINE, A.; ROZINAJOVÁ, V.

Originální název

Fully differentiable Lagrangian convolutional neural network for physics-informed precipitation nowcasting

Anglický název

Fully differentiable Lagrangian convolutional neural network for physics-informed precipitation nowcasting

Druh

Článek WoS

Originální abstrakt

This paper presents a convolutional neural network model for precipitation nowcasting that combines data-driven learning with physics-informed domain knowledge. We propose LUPIN, a Lagrangian Double U-Net for Physics-Informed Nowcasting, that draws from existing extrapolation-based nowcasting methods. It consists of a U-Net that dynamically produces mesoscale advection motion fields, a differentiable semi-Lagrangian extrapolation operator, and an advection-free U-Net capturing the growth and decay of precipitation over time. Using our approach, we successfully implement the Lagrangian convolutional neural network for precipitation nowcasting in a fully differentiable and GPU-accelerated manner. This allows for end-to-end training and inference, including the data-driven Lagrangian coordinate system transformation of the data at runtime. We evaluate the model and compare it with other related AI-based models both quantitatively and qualitatively in an extreme event case study. Based on our evaluation, LUPIN matches and even exceeds the performance of the chosen benchmarks, opening the door for other Lagrangian machine learning models.

Anglický abstrakt

This paper presents a convolutional neural network model for precipitation nowcasting that combines data-driven learning with physics-informed domain knowledge. We propose LUPIN, a Lagrangian Double U-Net for Physics-Informed Nowcasting, that draws from existing extrapolation-based nowcasting methods. It consists of a U-Net that dynamically produces mesoscale advection motion fields, a differentiable semi-Lagrangian extrapolation operator, and an advection-free U-Net capturing the growth and decay of precipitation over time. Using our approach, we successfully implement the Lagrangian convolutional neural network for precipitation nowcasting in a fully differentiable and GPU-accelerated manner. This allows for end-to-end training and inference, including the data-driven Lagrangian coordinate system transformation of the data at runtime. We evaluate the model and compare it with other related AI-based models both quantitatively and qualitatively in an extreme event case study. Based on our evaluation, LUPIN matches and even exceeds the performance of the chosen benchmarks, opening the door for other Lagrangian machine learning models.

Klíčová slova

Rain, Forecasting, Neural networks, Meteorological radar

Klíčová slova v angličtině

Rain, Forecasting, Neural networks, Meteorological radar

Autoři

PAVLÍK, P.; VÝBOH, M.; BOU EZZEDDINE, A.; ROZINAJOVÁ, V.

Rok RIV

2026

Vydáno

01.12.2025

Nakladatel

Elsevier

Periodikum

Applied computing and geosciences

Svazek

28

Číslo

December

Stát

Spojené království Velké Británie a Severního Irska

Strany od

1

Strany do

10

Strany počet

10

URL

https://www.sciencedirect.com/science/article/pii/S2590197425000783

Plný text v Digitální knihovně

http://hdl.handle.net/11012/256472

BibTex

@article{BUT200728,
  author="Peter {Pavlík} and  {} and Anna {Bou Ezzeddine} and Věra {Rozinajová}",
  title="Fully differentiable Lagrangian convolutional neural network for physics-informed precipitation nowcasting",
  journal="Applied computing and geosciences",
  year="2025",
  volume="28",
  number="December",
  pages="10",
  doi="10.1016/j.acags.2025.100296",
  url="https://www.sciencedirect.com/science/article/pii/S2590197425000783"
}