Infilling of missing rainfall radar data with a memory-assisted deep learning approach

<p>Incomplete spatiotemporal meteorological observations can result in misinterpretations of the current climate state, uncertainties in early warning systems, or inaccuracies in nowcasting models and can thereby pose significant challenges in hydrology research or similar applications. Tradit...

Full description

Saved in:
Bibliographic Details
Main Authors: J. Meuer, L. M. Bouwer, F. Kaspar, R. Lehmann, W. Karl, T. Ludwig, C. Kadow
Format: Article
Language:English
Published: Copernicus Publications 2025-08-01
Series:Hydrology and Earth System Sciences
Online Access:https://hess.copernicus.org/articles/29/3687/2025/hess-29-3687-2025.pdf
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849767979188748288
author J. Meuer
L. M. Bouwer
L. M. Bouwer
F. Kaspar
R. Lehmann
W. Karl
T. Ludwig
C. Kadow
author_facet J. Meuer
L. M. Bouwer
L. M. Bouwer
F. Kaspar
R. Lehmann
W. Karl
T. Ludwig
C. Kadow
author_sort J. Meuer
collection DOAJ
description <p>Incomplete spatiotemporal meteorological observations can result in misinterpretations of the current climate state, uncertainties in early warning systems, or inaccuracies in nowcasting models and can thereby pose significant challenges in hydrology research or similar applications. Traditional statistical methods for infilling missing precipitation data demand substantial computational resources and fail over large areas with sparse data – like temporary outages of weather radars. Although recent machine learning advancements have shown promise in addressing missing meteorological or satellite observations, they typically focus on spatial aspects, overlooking the complex spatiotemporal variability characteristic of precipitation, especially during extreme events. We propose a deep convolutional neural network enhanced with a memory component to better account for temporal changes in precipitation fields. This approach can analyse arbitrary sequences from before and/or after the incomplete observation of interest. Our model is trained and evaluated on the hourly RADKLIM dataset, which features 1 km resolution precipitation data derived from combined radar and weather stations across Germany. By infilling both artificial and actual data gaps of RADKLIM, we demonstrate the model's effectiveness, providing detailed insights into its capabilities during significant rainfall events, such as those in May 2012 and July 2021, including those responsible for the Ahrtal flood. This novel approach represents a step forward in hydrological applications, potentially improving the way we predict and manage water-related events by increasing the accuracy and reliability of precipitation data analysis.</p>
format Article
id doaj-art-265c6ecd1bd048d6b8559072da80d35c
institution DOAJ
issn 1027-5606
1607-7938
language English
publishDate 2025-08-01
publisher Copernicus Publications
record_format Article
series Hydrology and Earth System Sciences
spelling doaj-art-265c6ecd1bd048d6b8559072da80d35c2025-08-20T03:03:59ZengCopernicus PublicationsHydrology and Earth System Sciences1027-56061607-79382025-08-01293687370110.5194/hess-29-3687-2025Infilling of missing rainfall radar data with a memory-assisted deep learning approachJ. Meuer0L. M. Bouwer1L. M. Bouwer2F. Kaspar3R. Lehmann4W. Karl5T. Ludwig6C. Kadow7Data Analysis Department, German Climate Computing Center (DKRZ), Hamburg, GermanyClimate Service Center Germany (GERICS), Helmholtz-Zentrum Hereon, Hamburg, GermanyInstitute of Geography, University of Hamburg, Hamburg, GermanyHydrometeorology Department, Deutscher Wetterdienst (DWD), Offenbach, GermanyInstitute for Technical Informatics, Karlsruhe Institute of Technology (KIT), Karlsruhe, GermanyInstitute for Technical Informatics, Karlsruhe Institute of Technology (KIT), Karlsruhe, GermanyData Analysis Department, German Climate Computing Center (DKRZ), Hamburg, GermanyData Analysis Department, German Climate Computing Center (DKRZ), Hamburg, Germany<p>Incomplete spatiotemporal meteorological observations can result in misinterpretations of the current climate state, uncertainties in early warning systems, or inaccuracies in nowcasting models and can thereby pose significant challenges in hydrology research or similar applications. Traditional statistical methods for infilling missing precipitation data demand substantial computational resources and fail over large areas with sparse data – like temporary outages of weather radars. Although recent machine learning advancements have shown promise in addressing missing meteorological or satellite observations, they typically focus on spatial aspects, overlooking the complex spatiotemporal variability characteristic of precipitation, especially during extreme events. We propose a deep convolutional neural network enhanced with a memory component to better account for temporal changes in precipitation fields. This approach can analyse arbitrary sequences from before and/or after the incomplete observation of interest. Our model is trained and evaluated on the hourly RADKLIM dataset, which features 1 km resolution precipitation data derived from combined radar and weather stations across Germany. By infilling both artificial and actual data gaps of RADKLIM, we demonstrate the model's effectiveness, providing detailed insights into its capabilities during significant rainfall events, such as those in May 2012 and July 2021, including those responsible for the Ahrtal flood. This novel approach represents a step forward in hydrological applications, potentially improving the way we predict and manage water-related events by increasing the accuracy and reliability of precipitation data analysis.</p>https://hess.copernicus.org/articles/29/3687/2025/hess-29-3687-2025.pdf
spellingShingle J. Meuer
L. M. Bouwer
L. M. Bouwer
F. Kaspar
R. Lehmann
W. Karl
T. Ludwig
C. Kadow
Infilling of missing rainfall radar data with a memory-assisted deep learning approach
Hydrology and Earth System Sciences
title Infilling of missing rainfall radar data with a memory-assisted deep learning approach
title_full Infilling of missing rainfall radar data with a memory-assisted deep learning approach
title_fullStr Infilling of missing rainfall radar data with a memory-assisted deep learning approach
title_full_unstemmed Infilling of missing rainfall radar data with a memory-assisted deep learning approach
title_short Infilling of missing rainfall radar data with a memory-assisted deep learning approach
title_sort infilling of missing rainfall radar data with a memory assisted deep learning approach
url https://hess.copernicus.org/articles/29/3687/2025/hess-29-3687-2025.pdf
work_keys_str_mv AT jmeuer infillingofmissingrainfallradardatawithamemoryassisteddeeplearningapproach
AT lmbouwer infillingofmissingrainfallradardatawithamemoryassisteddeeplearningapproach
AT lmbouwer infillingofmissingrainfallradardatawithamemoryassisteddeeplearningapproach
AT fkaspar infillingofmissingrainfallradardatawithamemoryassisteddeeplearningapproach
AT rlehmann infillingofmissingrainfallradardatawithamemoryassisteddeeplearningapproach
AT wkarl infillingofmissingrainfallradardatawithamemoryassisteddeeplearningapproach
AT tludwig infillingofmissingrainfallradardatawithamemoryassisteddeeplearningapproach
AT ckadow infillingofmissingrainfallradardatawithamemoryassisteddeeplearningapproach