Building Tangent‐Linear and Adjoint Models for Data Assimilation With Neural Networks

Abstract We assess the ability of neural network emulators of physical parametrization schemes in numerical weather prediction models to aid in the construction of linearized models required by four‐dimensional variational (4D‐Var) data assimilation. Neural networks can be differentiated trivially,...

Full description

Bibliographic Details
Main Authors: Sam Hatfield, Matthew Chantry, Peter Dueben, Philippe Lopez, Alan Geer, Tim Palmer
Format: Article
Language:English
Published: American Geophysical Union (AGU) 2021-09-01
Series:Journal of Advances in Modeling Earth Systems
Subjects:
Online Access:https://doi.org/10.1029/2021MS002521
id doaj-541ee394ec4045ef8f348f8ed98f6e2e
record_format Article
spelling doaj-541ee394ec4045ef8f348f8ed98f6e2e2021-09-28T06:35:39ZengAmerican Geophysical Union (AGU)Journal of Advances in Modeling Earth Systems1942-24662021-09-01139n/an/a10.1029/2021MS002521Building Tangent‐Linear and Adjoint Models for Data Assimilation With Neural NetworksSam Hatfield0Matthew Chantry1Peter Dueben2Philippe Lopez3Alan Geer4Tim Palmer5European Centre for Medium‐Range Weather Forecasts Reading UKAtmospheric, Oceanic and Planetary Physics, University of Oxford Oxford UKEuropean Centre for Medium‐Range Weather Forecasts Reading UKEuropean Centre for Medium‐Range Weather Forecasts Reading UKEuropean Centre for Medium‐Range Weather Forecasts Reading UKAtmospheric, Oceanic and Planetary Physics, University of Oxford Oxford UKAbstract We assess the ability of neural network emulators of physical parametrization schemes in numerical weather prediction models to aid in the construction of linearized models required by four‐dimensional variational (4D‐Var) data assimilation. Neural networks can be differentiated trivially, and so if a physical parametrization scheme can be accurately emulated by a neural network then its tangent‐linear and adjoint versions can be obtained with minimal effort, compared with the standard paradigms of manual or automatic differentiation of the model code. Here we apply this idea by emulating the non‐orographic gravity wave drag parametrization scheme in an atmospheric model with a neural network, and deriving its tangent‐linear and adjoint models. We demonstrate that these neural network‐derived tangent‐linear and adjoint models not only pass the standard consistency tests but also can be used successfully to do 4D‐Var data assimilation. This technique holds the promise of significantly easing maintenance of tangent‐linear and adjoint codes in weather forecasting centers, if accurate neural network emulators can be constructed.https://doi.org/10.1029/2021MS002521neural networkdata assimilationtangent‐linearadjoint
collection DOAJ
language English
format Article
sources DOAJ
author Sam Hatfield
Matthew Chantry
Peter Dueben
Philippe Lopez
Alan Geer
Tim Palmer
spellingShingle Sam Hatfield
Matthew Chantry
Peter Dueben
Philippe Lopez
Alan Geer
Tim Palmer
Building Tangent‐Linear and Adjoint Models for Data Assimilation With Neural Networks
Journal of Advances in Modeling Earth Systems
neural network
data assimilation
tangent‐linear
adjoint
author_facet Sam Hatfield
Matthew Chantry
Peter Dueben
Philippe Lopez
Alan Geer
Tim Palmer
author_sort Sam Hatfield
title Building Tangent‐Linear and Adjoint Models for Data Assimilation With Neural Networks
title_short Building Tangent‐Linear and Adjoint Models for Data Assimilation With Neural Networks
title_full Building Tangent‐Linear and Adjoint Models for Data Assimilation With Neural Networks
title_fullStr Building Tangent‐Linear and Adjoint Models for Data Assimilation With Neural Networks
title_full_unstemmed Building Tangent‐Linear and Adjoint Models for Data Assimilation With Neural Networks
title_sort building tangent‐linear and adjoint models for data assimilation with neural networks
publisher American Geophysical Union (AGU)
series Journal of Advances in Modeling Earth Systems
issn 1942-2466
publishDate 2021-09-01
description Abstract We assess the ability of neural network emulators of physical parametrization schemes in numerical weather prediction models to aid in the construction of linearized models required by four‐dimensional variational (4D‐Var) data assimilation. Neural networks can be differentiated trivially, and so if a physical parametrization scheme can be accurately emulated by a neural network then its tangent‐linear and adjoint versions can be obtained with minimal effort, compared with the standard paradigms of manual or automatic differentiation of the model code. Here we apply this idea by emulating the non‐orographic gravity wave drag parametrization scheme in an atmospheric model with a neural network, and deriving its tangent‐linear and adjoint models. We demonstrate that these neural network‐derived tangent‐linear and adjoint models not only pass the standard consistency tests but also can be used successfully to do 4D‐Var data assimilation. This technique holds the promise of significantly easing maintenance of tangent‐linear and adjoint codes in weather forecasting centers, if accurate neural network emulators can be constructed.
topic neural network
data assimilation
tangent‐linear
adjoint
url https://doi.org/10.1029/2021MS002521
work_keys_str_mv AT samhatfield buildingtangentlinearandadjointmodelsfordataassimilationwithneuralnetworks
AT matthewchantry buildingtangentlinearandadjointmodelsfordataassimilationwithneuralnetworks
AT peterdueben buildingtangentlinearandadjointmodelsfordataassimilationwithneuralnetworks
AT philippelopez buildingtangentlinearandadjointmodelsfordataassimilationwithneuralnetworks
AT alangeer buildingtangentlinearandadjointmodelsfordataassimilationwithneuralnetworks
AT timpalmer buildingtangentlinearandadjointmodelsfordataassimilationwithneuralnetworks
_version_ 1716866398882365440