Variational Data Assimilation with a Learned Inverse Observation Operator

Thomas Frerix, Dmitrii Kochkov, Jamie Smith, Daniel Cremers, Michael Brenner, Stephan Hoyer
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:3449-3458, 2021.

Abstract

Variational data assimilation optimizes for an initial state of a dynamical system such that its evolution fits observational data. The physical model can subsequently be evolved into the future to make predictions. This principle is a cornerstone of large scale forecasting applications such as numerical weather prediction. As such, it is implemented in current operational systems of weather forecasting agencies across the globe. However, finding a good initial state poses a difficult optimization problem in part due to the non-invertible relationship between physical states and their corresponding observations. We learn a mapping from observational data to physical states and show how it can be used to improve optimizability. We employ this mapping in two ways: to better initialize the non-convex optimization problem, and to reformulate the objective function in better behaved physics space instead of observation space. Our experimental results for the Lorenz96 model and a two-dimensional turbulent fluid flow demonstrate that this procedure significantly improves forecast quality for chaotic systems.

Cite this Paper

BibTeX
@InProceedings{pmlr-v139-frerix21a, title = {Variational Data Assimilation with a Learned Inverse Observation Operator}, author = {Frerix, Thomas and Kochkov, Dmitrii and Smith, Jamie and Cremers, Daniel and Brenner, Michael and Hoyer, Stephan}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {3449--3458}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/frerix21a/frerix21a.pdf}, url = {https://proceedings.mlr.press/v139/frerix21a.html}, abstract = {Variational data assimilation optimizes for an initial state of a dynamical system such that its evolution fits observational data. The physical model can subsequently be evolved into the future to make predictions. This principle is a cornerstone of large scale forecasting applications such as numerical weather prediction. As such, it is implemented in current operational systems of weather forecasting agencies across the globe. However, finding a good initial state poses a difficult optimization problem in part due to the non-invertible relationship between physical states and their corresponding observations. We learn a mapping from observational data to physical states and show how it can be used to improve optimizability. We employ this mapping in two ways: to better initialize the non-convex optimization problem, and to reformulate the objective function in better behaved physics space instead of observation space. Our experimental results for the Lorenz96 model and a two-dimensional turbulent fluid flow demonstrate that this procedure significantly improves forecast quality for chaotic systems.} }
Endnote
%0 Conference Paper %T Variational Data Assimilation with a Learned Inverse Observation Operator %A Thomas Frerix %A Dmitrii Kochkov %A Jamie Smith %A Daniel Cremers %A Michael Brenner %A Stephan Hoyer %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-frerix21a %I PMLR %P 3449--3458 %U https://proceedings.mlr.press/v139/frerix21a.html %V 139 %X Variational data assimilation optimizes for an initial state of a dynamical system such that its evolution fits observational data. The physical model can subsequently be evolved into the future to make predictions. This principle is a cornerstone of large scale forecasting applications such as numerical weather prediction. As such, it is implemented in current operational systems of weather forecasting agencies across the globe. However, finding a good initial state poses a difficult optimization problem in part due to the non-invertible relationship between physical states and their corresponding observations. We learn a mapping from observational data to physical states and show how it can be used to improve optimizability. We employ this mapping in two ways: to better initialize the non-convex optimization problem, and to reformulate the objective function in better behaved physics space instead of observation space. Our experimental results for the Lorenz96 model and a two-dimensional turbulent fluid flow demonstrate that this procedure significantly improves forecast quality for chaotic systems.
APA
Frerix, T., Kochkov, D., Smith, J., Cremers, D., Brenner, M. & Hoyer, S.. (2021). Variational Data Assimilation with a Learned Inverse Observation Operator. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:3449-3458 Available from https://proceedings.mlr.press/v139/frerix21a.html.

Related Material