• Corpus ID: 244130419

Adjoint-Matching Neural Network Surrogates for Fast 4D-Var Data Assimilation

  title={Adjoint-Matching Neural Network Surrogates for Fast 4D-Var Data Assimilation},
  author={Austin Chennault and Andrey A. Popov and Amit N. Subrahmanya and Rachel Cooper and Anuj Karpatne and Adrian Sandu},
The data assimilation procedures used in many operational numerical weather forecasting systems are based around variants of the 4D-Var algorithm. The cost of solving the 4D-Var problem is dominated by the cost of forward and adjoint evaluations of the physical model. This motivates their substitution by fast, approximate surrogate models. Neural networks offer a promising approach for the data-driven creation of surrogate models. The accuracy of the surrogate 4D-Var problem’s solution has been… 

Figures from this paper

Equation-free surrogate modeling of geophysical flows at the intersection of machine learning and data assimilation

This work introduces an end-to-end non-intrusive reduced-order modeling (NIROM) framework equipped with contributions in modal decomposition, time series prediction, optimal sensor placement, and sequential data assimilation and indicates that the NIROM is stable for long-term forecasting and can model dynamics of SST with a reasonable level of accuracy.

Efficient high-dimensional variational data assimilation with machine-learned reduced-order models

The overall formulation is denoted AIEADA (Artificial Intelligence Emulator-Assisted Data Assimilation), indicating 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.

CSL-TR-22-2 August 22 , 2022

This work proposes Control Physics-Informed Neural Networks (Control PINNs) that simultaneously solve for a given system state, and for the optimal control signal, in a one-stage framework that conforms to the underlying physical laws.

CSL-TR-22-2 May 9 , 2022

  • 2022



A comparison of combined data assimilation and machine learning methods for offline and online model error correction

Variational chemical data assimilation with approximate adjoints

A data-driven non-linear assimilation framework with neural networks

This paper develops assimilation methods by building powerful surrogates that emulate the evolution of the model observables of the dynamical system to efficiently perform assimilation on the reduced model and demonstrates on a chaotic test case that uncertainty in initial condition is accurately captured by the surrogate.

Adjoint sensitivity analysis of regional air quality models

An adjoint sensitivity analysis and 4D-Var data assimilation study of Texas air quality

Combining data assimilation and machine learning to emulate a dynamical model from sparse and noisy observations: a case study with the Lorenz 96 model

The synergy demonstrated with a low-dimensional system is encouraging for more sophisticated dynamics and motivates further investigation to merge data assimilation and machine learning.

Deep Data Assimilation: Integrating Deep Learning with Data Assimilation

The DDA technology is applied to two different applications: the Double integral mass dot system and the Lorenz system, and it is proved that the DDA approach implies a reduction of the model error, which decreases at each iteration.

Multifidelity Ensemble Kalman Filtering using surrogate models defined by Physics-Informed Autoencoders

Optimal nonlinear projection and interpolation operators are obtained by appropriately trained physics-informed autoencoders, and this approach allows to construct reduced order surrogate models with less error than conventional linear methods.

POD/DEIM Strategies for reduced data assimilation systems

Reduced order data assimilation requires low-rank surrogate models that accurately represent sub-grid-scale processes and highly non-linear observation operators to be minimized in the inner loop.

A Multifidelity Ensemble Kalman Filter with Reduced Order Control Variates

Numerical results show that the two-fidelity MFEnKF provides better analyses than existing EnKF algorithms at comparable or reduced computational costs.