Efficient high-dimensional variational data assimilation with machine-learned reduced-order models
<p>Data assimilation (DA) in geophysical sciences remains the cornerstone of robust forecasts from numerical models. Indeed, DA plays a crucial role in the quality of numerical weather prediction and is a crucial building block that has allowed dramatic improvements in weather forecasting over...
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Copernicus Publications
2022-05-01
|
| Series: | Geoscientific Model Development |
| Online Access: | https://gmd.copernicus.org/articles/15/3433/2022/gmd-15-3433-2022.pdf |
| _version_ | 1811291303953039360 |
|---|---|
| author | R. Maulik V. Rao J. Wang G. Mengaldo E. Constantinescu B. Lusch P. Balaprakash I. Foster R. Kotamarthi |
| author_facet | R. Maulik V. Rao J. Wang G. Mengaldo E. Constantinescu B. Lusch P. Balaprakash I. Foster R. Kotamarthi |
| author_sort | R. Maulik |
| collection | DOAJ |
| description | <p>Data assimilation (DA) in geophysical sciences remains the cornerstone of robust forecasts from numerical models. Indeed, DA plays a crucial role in the quality of numerical weather prediction and is a crucial building block that has allowed dramatic improvements in weather forecasting over the past few decades. DA is commonly framed in a variational setting, where one solves an optimization problem within a Bayesian formulation using raw model forecasts as a prior and observations as likelihood. This leads to a DA objective function that needs to be minimized, where the decision variables are the initial conditions specified to the model. In traditional DA, the forward model is numerically and computationally expensive. Here we replace the forward model with a low-dimensional, data-driven, and differentiable emulator. Consequently, gradients of our DA objective function with respect to the decision variables are obtained rapidly via automatic differentiation. We demonstrate our approach by performing an emulator-assisted DA forecast of geopotential height. Our results indicate that emulator-assisted DA is faster than traditional equation-based DA forecasts by 4 orders of magnitude, allowing computations to be performed on a workstation rather than a dedicated high-performance computer. In addition, we describe accuracy benefits of emulator-assisted DA when compared to simply using the emulator for forecasting (i.e., without DA). Our overall formulation is denoted AIEADA (Artificial Intelligence Emulator-Assisted Data Assimilation).</p> |
| first_indexed | 2024-04-13T04:27:21Z |
| format | Article |
| id | doaj.art-0aff770152eb4f1ea9aeb288f709b8db |
| institution | Directory Open Access Journal |
| issn | 1991-959X 1991-9603 |
| language | English |
| last_indexed | 2024-04-13T04:27:21Z |
| publishDate | 2022-05-01 |
| publisher | Copernicus Publications |
| record_format | Article |
| series | Geoscientific Model Development |
| spelling | doaj.art-0aff770152eb4f1ea9aeb288f709b8db2022-12-22T03:02:28ZengCopernicus PublicationsGeoscientific Model Development1991-959X1991-96032022-05-01153433344510.5194/gmd-15-3433-2022Efficient high-dimensional variational data assimilation with machine-learned reduced-order modelsR. Maulik0V. Rao1J. Wang2G. Mengaldo3E. Constantinescu4B. Lusch5P. Balaprakash6I. Foster7R. Kotamarthi8240, Argonne National Laboratory, Lemont, IL 60439, USA240, Argonne National Laboratory, Lemont, IL 60439, USA240, Argonne National Laboratory, Lemont, IL 60439, USADepartment of Mechanical Engineering, National University of Singapore, Block EA, #07-08, 9 Engineering Drive 1, Singapore240, Argonne National Laboratory, Lemont, IL 60439, USA240, Argonne National Laboratory, Lemont, IL 60439, USA240, Argonne National Laboratory, Lemont, IL 60439, USA240, Argonne National Laboratory, Lemont, IL 60439, USA240, Argonne National Laboratory, Lemont, IL 60439, USA<p>Data assimilation (DA) in geophysical sciences remains the cornerstone of robust forecasts from numerical models. Indeed, DA plays a crucial role in the quality of numerical weather prediction and is a crucial building block that has allowed dramatic improvements in weather forecasting over the past few decades. DA is commonly framed in a variational setting, where one solves an optimization problem within a Bayesian formulation using raw model forecasts as a prior and observations as likelihood. This leads to a DA objective function that needs to be minimized, where the decision variables are the initial conditions specified to the model. In traditional DA, the forward model is numerically and computationally expensive. Here we replace the forward model with a low-dimensional, data-driven, and differentiable emulator. Consequently, gradients of our DA objective function with respect to the decision variables are obtained rapidly via automatic differentiation. We demonstrate our approach by performing an emulator-assisted DA forecast of geopotential height. Our results indicate that emulator-assisted DA is faster than traditional equation-based DA forecasts by 4 orders of magnitude, allowing computations to be performed on a workstation rather than a dedicated high-performance computer. In addition, we describe accuracy benefits of emulator-assisted DA when compared to simply using the emulator for forecasting (i.e., without DA). Our overall formulation is denoted AIEADA (Artificial Intelligence Emulator-Assisted Data Assimilation).</p>https://gmd.copernicus.org/articles/15/3433/2022/gmd-15-3433-2022.pdf |
| spellingShingle | R. Maulik V. Rao J. Wang G. Mengaldo E. Constantinescu B. Lusch P. Balaprakash I. Foster R. Kotamarthi Efficient high-dimensional variational data assimilation with machine-learned reduced-order models Geoscientific Model Development |
| title | Efficient high-dimensional variational data assimilation with machine-learned reduced-order models |
| title_full | Efficient high-dimensional variational data assimilation with machine-learned reduced-order models |
| title_fullStr | Efficient high-dimensional variational data assimilation with machine-learned reduced-order models |
| title_full_unstemmed | Efficient high-dimensional variational data assimilation with machine-learned reduced-order models |
| title_short | Efficient high-dimensional variational data assimilation with machine-learned reduced-order models |
| title_sort | efficient high dimensional variational data assimilation with machine learned reduced order models |
| url | https://gmd.copernicus.org/articles/15/3433/2022/gmd-15-3433-2022.pdf |
| work_keys_str_mv | AT rmaulik efficienthighdimensionalvariationaldataassimilationwithmachinelearnedreducedordermodels AT vrao efficienthighdimensionalvariationaldataassimilationwithmachinelearnedreducedordermodels AT jwang efficienthighdimensionalvariationaldataassimilationwithmachinelearnedreducedordermodels AT gmengaldo efficienthighdimensionalvariationaldataassimilationwithmachinelearnedreducedordermodels AT econstantinescu efficienthighdimensionalvariationaldataassimilationwithmachinelearnedreducedordermodels AT blusch efficienthighdimensionalvariationaldataassimilationwithmachinelearnedreducedordermodels AT pbalaprakash efficienthighdimensionalvariationaldataassimilationwithmachinelearnedreducedordermodels AT ifoster efficienthighdimensionalvariationaldataassimilationwithmachinelearnedreducedordermodels AT rkotamarthi efficienthighdimensionalvariationaldataassimilationwithmachinelearnedreducedordermodels |