Sarah Williamson
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Sarah Williamson has a total of 29 co-authors. Here are the top co-authors by number of shared publications:
- Carl Wunsch
Department of Earth and Planetary Sciences, Harvard University, Cambridge, Massachusetts
6 shared publications Patrick Heimbach
Oden Institute for Computational Engineering and Sciences, University of Texas at Austin, Austin, TX, USA
5 shared publications- Patrick Heimbach
The University of Texas at Austin
4 shared publications - Sri Hari Krishna Narayanan
Argonne National Lab
4 shared publications Michel Schanen
Argonne National Laboratory, Lemont, IL, Lemont, Illinois, United States
3 shared publications- Nora Loose
University of Colorado Boulder
3 shared publications
Latest Publications
- Open Access
- 12 November 2026
- Published by Wiley
An online-learned eddy parameterization in a differentiable shallow water model
Ocean general circulation models rely on parameterizations to approximate subgrid-scale processes. While parameterizations are traditionally physics-based, machine learning has recently provided an alternative, data-driven approach for developing parameterizations. However, most machine learning-based parameterizations are learned offline due to the lack of differentiable general circulation models. Here, we train an eddy momentum parameterization represented by a neural network online in an idealized barotropic gyre model. We develop the model as fully differentiable in Julia with the automatic differentiation tool Enzyme. The network weights are trained using a loss function that has the same form as in adjoint-based data assimilation, with synthetic observations generated by a fine-resolution simulation. A range of diagnostics evaluate the performance of the parameterized coarse-resolution model. We find improved energetics, physically realistic dynamics, and numerical stability for multi-decadal roll-outs. Importantly, the online-learned parameterization absorbs artifacts from the numerical discretization, such as the time-stepping scheme used, which is invisible to offline-learned parameterizations. We suggest that this is key to improved numerical stability of machine-learned parameterizations. Our research suggests that online-trained neural network parameterizations can be an effective solution to the subgrid-scale parameterization and model calibration problem.
- 12 November 2026
- Published by Wiley
Open AccessAn online-learned eddy parameterization in a differentiable shallow water model
Ocean general circulation models rely on parameterizations to approximate subgrid-scale processes. While parameterizations are traditionally physics-based, machine learning has recently provided an alternative, data-driven approach for developing parameterizations. However, most machine learning-based parameterizations are learned offline due to the lack of differentiable general circulation models. Here, we train an eddy momentum parameterization represented by a neural network online in an idealized barotropic gyre model. We develop the model as fully differentiable in Julia with the automatic differentiation tool Enzyme. The network weights are trained using a loss function that has the same form as in adjoint-based data assimilation, with synthetic observations generated by a fine-resolution simulation. A range of diagnostics evaluate the performance of the parameterized coarse-resolution model. We find improved energetics, physically realistic dynamics, and numerical stability for multi-decadal roll-outs. Importantly, the online-learned parameterization absorbs artifacts from the numerical discretization, such as the time-stepping scheme used, which is invisible to offline-learned parameterizations. We suggest that this is key to improved numerical stability of machine-learned parameterizations. Our research suggests that online-trained neural network parameterizations can be an effective solution to the subgrid-scale parameterization and model calibration problem.
- Open Access
- 18 May 2026
- Published by American Geophysical Union (AGU) in Journal of Advances in Modeling Earth Systems
- Vol. 18 (5)
DJ4Earth: Differentiable, and Performance‐Portable Earth System Modeling via Program Transformations
Differentiable Earth system models (ESMs) enable powerful applications such as sensitivity analysis, gradient‐based calibration, state estimation, boundary flux inversions, uncertainty quantification, and online machine learning. Reverse‐mode automatic differentiation (AD) efficiently provides gradients for such tasks, yet models have rarely included this capability because of complex, bespoke numerical algorithms. As part of the Differentiable programming in Julia for Earth system modeling (DJ4Earth) initiative, we present improved capabilities of the AD tool Enzyme.jl and the new compiler transpilation tool Reactant.jl, augmented by sophisticated checkpointing algorithms, which, together make general‐purpose AD tractable and efficient for full‐fledged ESM components written in Julia. Operating at the low‐level virtual machine intermediate representation or multi‐level intermediate representation compiler levels, these frameworks support mutable memory, custom kernels, and compiler optimizations before and after differentiation. Julia‐specific challenges related to just‐in‐time compilation and garbage collection are handled efficiently. Reactant further enables automatic performance portability across central processing units, graphics processing units, and tensor processing units, facilitating use of emerging AI‐customized high‐performance computing architectures. We demonstrate these frameworks on four Julia‐based ESM components featuring diverse spatial discretizations and numerical algorithms: the rotating‐sphere shallow water model ShallowWaters.jl, the finite‐volume ocean model Oceananigans.jl, the finite‐element ice sheet model DJUICE.jl, and the spectral atmospheric model SpeedyWeather.jl. Across these ESM components, our tools compute efficient and correct gradients. These results establish a foundation for differentiable, high‐performance and performance‐portable ESMs that can integrate neural networks for unresolved processes, trained online, enabling next‐generation hybrid physics–machine learning ESMs constrained by physical dynamics and observations.
- 18 May 2026
- Published by American Geophysical Union (AGU) in Journal of Advances in Modeling Earth Systems
- Vol. 18 (5)
Open AccessDJ4Earth: Differentiable, and Performance‐Portable Earth System Modeling via Program Transformations
Differentiable Earth system models (ESMs) enable powerful applications such as sensitivity analysis, gradient‐based calibration, state estimation, boundary flux inversions, uncertainty quantification, and online machine learning. Reverse‐mode automatic differentiation (AD) efficiently provides gradients for such tasks, yet models have rarely included this capability because of complex, bespoke numerical algorithms. As part of the Differentiable programming in Julia for Earth system modeling (DJ4Earth) initiative, we present improved capabilities of the AD tool Enzyme.jl and the new compiler transpilation tool Reactant.jl, augmented by sophisticated checkpointing algorithms, which, together make general‐purpose AD tractable and efficient for full‐fledged ESM components written in Julia. Operating at the low‐level virtual machine intermediate representation or multi‐level intermediate representation compiler levels, these frameworks support mutable memory, custom kernels, and compiler optimizations before and after differentiation. Julia‐specific challenges related to just‐in‐time compilation and garbage collection are handled efficiently. Reactant further enables automatic performance portability across central processing units, graphics processing units, and tensor processing units, facilitating use of emerging AI‐customized high‐performance computing architectures. We demonstrate these frameworks on four Julia‐based ESM components featuring diverse spatial discretizations and numerical algorithms: the rotating‐sphere shallow water model ShallowWaters.jl, the finite‐volume ocean model Oceananigans.jl, the finite‐element ice sheet model DJUICE.jl, and the spectral atmospheric model SpeedyWeather.jl. Across these ESM components, our tools compute efficient and correct gradients. These results establish a foundation for differentiable, high‐performance and performance‐portable ESMs that can integrate neural networks for unresolved processes, trained online, enabling next‐generation hybrid physics–machine learning ESMs constrained by physical dynamics and observations.
- Open Access
- 30 April 2026
- Published by Wiley
DJ4Earth: Differentiable, and Performance-portable Earth System Modeling via Program Transformations
Differentiable Earth system models (ESMs) enable powerful applications such as sensitivity analysis, gradient-based calibration, state estimation, boundary flux inversions, uncertainty quantification, and online machine learning. Reverse-mode automatic differentiation (AD) efficiently provides gradients for such tasks, yet models have rarely included this capability because of complex, bespoke numerical algorithms. As part of the \textit{Differentiable programming in Julia for Earth system modeling (DJ4Earth)} initiative, we present improved capabilities of the AD tool Enzyme.jl and the new compiler transpilation tool Reactant.jl, augmented by sophisticated checkpointing algorithms, which, together make general-purpose AD tractable and efficient for full-fledged ESM components written in Julia. Operating at the Low-Level Virtual Machine (LLVM) intermediate representation or Multi-Level Intermediate Representation (MLIR) compiler levels, these frameworks support mutable memory, custom kernels, and compiler optimizations before and after differentiation. Julia-specific challenges related to just-in-time compilation and garbage collection are handled efficiently. Reactant further enables automatic performance portability across CPUs, GPUs, and TPUs, facilitating use of emerging AI-customized high-performance computing architectures. We demonstrate these frameworks on four Julia-based ESM components featuring diverse spatial discretizations and numerical algorithms: (i) the rotating-sphere shallow water model ShallowWaters.jl, (ii) the finite-volume ocean model Oceananigans.jl, (iii) the finite-element ice sheet model DJUICE.jl, and (iv) the spectral atmospheric model SpeedyWeather.jl. Across these ESM components, our tools compute efficient and correct gradients. These results establish a foundation for differentiable, high-performance and performance-portable ESMs that can integrate neural networks for unresolved processes, trained online, enabling next-generation hybrid physics--machine learning ESMs constrained by physical dynamics and observations.
- 30 April 2026
- Published by Wiley
Open AccessDJ4Earth: Differentiable, and Performance-portable Earth System Modeling via Program Transformations
Differentiable Earth system models (ESMs) enable powerful applications such as sensitivity analysis, gradient-based calibration, state estimation, boundary flux inversions, uncertainty quantification, and online machine learning. Reverse-mode automatic differentiation (AD) efficiently provides gradients for such tasks, yet models have rarely included this capability because of complex, bespoke numerical algorithms. As part of the \textit{Differentiable programming in Julia for Earth system modeling (DJ4Earth)} initiative, we present improved capabilities of the AD tool Enzyme.jl and the new compiler transpilation tool Reactant.jl, augmented by sophisticated checkpointing algorithms, which, together make general-purpose AD tractable and efficient for full-fledged ESM components written in Julia. Operating at the Low-Level Virtual Machine (LLVM) intermediate representation or Multi-Level Intermediate Representation (MLIR) compiler levels, these frameworks support mutable memory, custom kernels, and compiler optimizations before and after differentiation. Julia-specific challenges related to just-in-time compilation and garbage collection are handled efficiently. Reactant further enables automatic performance portability across CPUs, GPUs, and TPUs, facilitating use of emerging AI-customized high-performance computing architectures. We demonstrate these frameworks on four Julia-based ESM components featuring diverse spatial discretizations and numerical algorithms: (i) the rotating-sphere shallow water model ShallowWaters.jl, (ii) the finite-volume ocean model Oceananigans.jl, (iii) the finite-element ice sheet model DJUICE.jl, and (iv) the spectral atmospheric model SpeedyWeather.jl. Across these ESM components, our tools compute efficient and correct gradients. These results establish a foundation for differentiable, high-performance and performance-portable ESMs that can integrate neural networks for unresolved processes, trained online, enabling next-generation hybrid physics--machine learning ESMs constrained by physical dynamics and observations.
- Open Access
- 21 August 2023
- Published by Copernicus GmbH in Ocean Science
- Vol. 19 (4) , 1253-1275
Potential artifacts in conservation laws and invariants inferred from sequential state estimation
In sequential estimation methods often used in oceanic and general climate calculations of the state and of forecasts, observations act mathematically and statistically as source or sink terms in conservation equations for heat, salt, mass, and momentum. These artificial terms obscure the inference of the system's variability or secular changes. Furthermore, for the purposes of calculating changes in important functions of state variables such as total mass and energy or volumetric current transports, results of both filter and smoother-based estimates are sensitive to misrepresentation of a large variety of parameters, including initial conditions, prior uncertainty covariances, and systematic and random errors in observations. Here, toy models of a coupled mass–spring oscillator system and of a barotropic Rossby wave system are used to demonstrate many of the issues that arise from such misrepresentations. Results from Kalman filter estimates and those from finite interval smoothing are analyzed. In the filter (and prediction) problem, entry of data leads to violation of conservation and other invariant rules. A finite interval smoothing method restores the conservation rules, but uncertainties in all such estimation results remain. Convincing trend and other time-dependent determinations in “reanalysis-like” estimates require a full understanding of models, observations, and underlying error structures. Application of smoother-type methods that are designed for optimal reconstruction purposes alleviate some of the issues.
- 21 August 2023
- Published by Copernicus GmbH in Ocean Science
- Vol. 19 (4) , 1253-1275
Open AccessPotential artifacts in conservation laws and invariants inferred from sequential state estimation
In sequential estimation methods often used in oceanic and general climate calculations of the state and of forecasts, observations act mathematically and statistically as source or sink terms in conservation equations for heat, salt, mass, and momentum. These artificial terms obscure the inference of the system's variability or secular changes. Furthermore, for the purposes of calculating changes in important functions of state variables such as total mass and energy or volumetric current transports, results of both filter and smoother-based estimates are sensitive to misrepresentation of a large variety of parameters, including initial conditions, prior uncertainty covariances, and systematic and random errors in observations. Here, toy models of a coupled mass–spring oscillator system and of a barotropic Rossby wave system are used to demonstrate many of the issues that arise from such misrepresentations. Results from Kalman filter estimates and those from finite interval smoothing are analyzed. In the filter (and prediction) problem, entry of data leads to violation of conservation and other invariant rules. A finite interval smoothing method restores the conservation rules, but uncertainties in all such estimation results remain. Convincing trend and other time-dependent determinations in “reanalysis-like” estimates require a full understanding of models, observations, and underlying error structures. Application of smoother-type methods that are designed for optimal reconstruction purposes alleviate some of the issues.
- 26 June 2023
- Published by Springer Nature
- p. 483-497
Transparent Checkpointing for Automatic Differentiation of Program Loops Through Expression Transformations
- 26 June 2023
- Published by Springer Nature
- p. 483-497
Transparent Checkpointing for Automatic Differentiation of Program Loops Through Expression Transformations
- Open Access
- 25 May 2023
- Published by Copernicus GmbH
- Vol. 2023, 1-34
Reply on RC1
In sequential estimation methods often used in oceanic and general climate calculations of the state and of forecasts, observations act mathematically and statistically as forcings. For purposes of calculating changes in important functions of state variables such as total mass and energy, or volumetric current transports, results are sensitive to mis-representation of a large variety of parameters, including initial conditions, prior uncertainty covariances, and systematic and random errors in observations. Here toy models of a mass-spring oscillator and of a barotropic Rossby-wave equation are used to demonstrate many of the issues. Results from Kalman-filter estimates, and those from finite interval smoothing are analyzed. In the filter (and prediction) problem, entry of data leads to violation of conservation and other invariant rules. A finite interval smoothing method restores the conservation rules, but uncertainties in all such estimation results remain. Convincing trend and other time-dependent determinations in "reanalysis" -like estimates require a full understanding of both models and observations.
- 25 May 2023
- Published by Copernicus GmbH
- Vol. 2023, 1-34
Open AccessReply on RC1
In sequential estimation methods often used in oceanic and general climate calculations of the state and of forecasts, observations act mathematically and statistically as forcings. For purposes of calculating changes in important functions of state variables such as total mass and energy, or volumetric current transports, results are sensitive to mis-representation of a large variety of parameters, including initial conditions, prior uncertainty covariances, and systematic and random errors in observations. Here toy models of a mass-spring oscillator and of a barotropic Rossby-wave equation are used to demonstrate many of the issues. Results from Kalman-filter estimates, and those from finite interval smoothing are analyzed. In the filter (and prediction) problem, entry of data leads to violation of conservation and other invariant rules. A finite interval smoothing method restores the conservation rules, but uncertainties in all such estimation results remain. Convincing trend and other time-dependent determinations in "reanalysis" -like estimates require a full understanding of both models and observations.
- Open Access
- 25 May 2023
- Published by Copernicus GmbH
- Vol. 2023, 1-34
Reply on RC2
In sequential estimation methods often used in oceanic and general climate calculations of the state and of forecasts, observations act mathematically and statistically as forcings. For purposes of calculating changes in important functions of state variables such as total mass and energy, or volumetric current transports, results are sensitive to mis-representation of a large variety of parameters, including initial conditions, prior uncertainty covariances, and systematic and random errors in observations. Here toy models of a mass-spring oscillator and of a barotropic Rossby-wave equation are used to demonstrate many of the issues. Results from Kalman-filter estimates, and those from finite interval smoothing are analyzed. In the filter (and prediction) problem, entry of data leads to violation of conservation and other invariant rules. A finite interval smoothing method restores the conservation rules, but uncertainties in all such estimation results remain. Convincing trend and other time-dependent determinations in "reanalysis" -like estimates require a full understanding of both models and observations.
- 25 May 2023
- Published by Copernicus GmbH
- Vol. 2023, 1-34
Open AccessReply on RC2
In sequential estimation methods often used in oceanic and general climate calculations of the state and of forecasts, observations act mathematically and statistically as forcings. For purposes of calculating changes in important functions of state variables such as total mass and energy, or volumetric current transports, results are sensitive to mis-representation of a large variety of parameters, including initial conditions, prior uncertainty covariances, and systematic and random errors in observations. Here toy models of a mass-spring oscillator and of a barotropic Rossby-wave equation are used to demonstrate many of the issues. Results from Kalman-filter estimates, and those from finite interval smoothing are analyzed. In the filter (and prediction) problem, entry of data leads to violation of conservation and other invariant rules. A finite interval smoothing method restores the conservation rules, but uncertainties in all such estimation results remain. Convincing trend and other time-dependent determinations in "reanalysis" -like estimates require a full understanding of both models and observations.
- Open Access
- 6 May 2023
- Published by Copernicus GmbH
- Vol. 2023, 1-34
Comment on egusphere-2023-140
In sequential estimation methods often used in oceanic and general climate calculations of the state and of forecasts, observations act mathematically and statistically as forcings. For purposes of calculating changes in important functions of state variables such as total mass and energy, or volumetric current transports, results are sensitive to mis-representation of a large variety of parameters, including initial conditions, prior uncertainty covariances, and systematic and random errors in observations. Here toy models of a mass-spring oscillator and of a barotropic Rossby-wave equation are used to demonstrate many of the issues. Results from Kalman-filter estimates, and those from finite interval smoothing are analyzed. In the filter (and prediction) problem, entry of data leads to violation of conservation and other invariant rules. A finite interval smoothing method restores the conservation rules, but uncertainties in all such estimation results remain. Convincing trend and other time-dependent determinations in "reanalysis" -like estimates require a full understanding of both models and observations.
- 6 May 2023
- Published by Copernicus GmbH
- Vol. 2023, 1-34
Open AccessComment on egusphere-2023-140
In sequential estimation methods often used in oceanic and general climate calculations of the state and of forecasts, observations act mathematically and statistically as forcings. For purposes of calculating changes in important functions of state variables such as total mass and energy, or volumetric current transports, results are sensitive to mis-representation of a large variety of parameters, including initial conditions, prior uncertainty covariances, and systematic and random errors in observations. Here toy models of a mass-spring oscillator and of a barotropic Rossby-wave equation are used to demonstrate many of the issues. Results from Kalman-filter estimates, and those from finite interval smoothing are analyzed. In the filter (and prediction) problem, entry of data leads to violation of conservation and other invariant rules. A finite interval smoothing method restores the conservation rules, but uncertainties in all such estimation results remain. Convincing trend and other time-dependent determinations in "reanalysis" -like estimates require a full understanding of both models and observations.
- Open Access
- 30 March 2023
- Published by Copernicus GmbH
- Vol. 2023, 1-34
Comment on egusphere-2023-140
In sequential estimation methods often used in oceanic and general climate calculations of the state and of forecasts, observations act mathematically and statistically as forcings. For purposes of calculating changes in important functions of state variables such as total mass and energy, or volumetric current transports, results are sensitive to mis-representation of a large variety of parameters, including initial conditions, prior uncertainty covariances, and systematic and random errors in observations. Here toy models of a mass-spring oscillator and of a barotropic Rossby-wave equation are used to demonstrate many of the issues. Results from Kalman-filter estimates, and those from finite interval smoothing are analyzed. In the filter (and prediction) problem, entry of data leads to violation of conservation and other invariant rules. A finite interval smoothing method restores the conservation rules, but uncertainties in all such estimation results remain. Convincing trend and other time-dependent determinations in "reanalysis" -like estimates require a full understanding of both models and observations.
- 30 March 2023
- Published by Copernicus GmbH
- Vol. 2023, 1-34
Open AccessComment on egusphere-2023-140
In sequential estimation methods often used in oceanic and general climate calculations of the state and of forecasts, observations act mathematically and statistically as forcings. For purposes of calculating changes in important functions of state variables such as total mass and energy, or volumetric current transports, results are sensitive to mis-representation of a large variety of parameters, including initial conditions, prior uncertainty covariances, and systematic and random errors in observations. Here toy models of a mass-spring oscillator and of a barotropic Rossby-wave equation are used to demonstrate many of the issues. Results from Kalman-filter estimates, and those from finite interval smoothing are analyzed. In the filter (and prediction) problem, entry of data leads to violation of conservation and other invariant rules. A finite interval smoothing method restores the conservation rules, but uncertainties in all such estimation results remain. Convincing trend and other time-dependent determinations in "reanalysis" -like estimates require a full understanding of both models and observations.
- Open Access
- 21 February 2023
- Published by Copernicus GmbH in EGUsphere
- Vol. 2023, 1-34
Potential Artifacts in Conservation Laws and Invariants Inferred from Sequential State Estimation
In sequential estimation methods often used in oceanic and general climate calculations of the state and of forecasts, observations act mathematically and statistically as forcings. For purposes of calculating changes in important functions of state variables such as total mass and energy, or volumetric current transports, results are sensitive to mis-representation of a large variety of parameters, including initial conditions, prior uncertainty covariances, and systematic and random errors in observations. Here toy models of a mass-spring oscillator and of a barotropic Rossby-wave equation are used to demonstrate many of the issues. Results from Kalman-filter estimates, and those from finite interval smoothing are analyzed. In the filter (and prediction) problem, entry of data leads to violation of conservation and other invariant rules. A finite interval smoothing method restores the conservation rules, but uncertainties in all such estimation results remain. Convincing trend and other time-dependent determinations in "reanalysis" -like estimates require a full understanding of both models and observations.
- 21 February 2023
- Published by Copernicus GmbH in EGUsphere
- Vol. 2023, 1-34
Open AccessPotential Artifacts in Conservation Laws and Invariants Inferred from Sequential State Estimation
In sequential estimation methods often used in oceanic and general climate calculations of the state and of forecasts, observations act mathematically and statistically as forcings. For purposes of calculating changes in important functions of state variables such as total mass and energy, or volumetric current transports, results are sensitive to mis-representation of a large variety of parameters, including initial conditions, prior uncertainty covariances, and systematic and random errors in observations. Here toy models of a mass-spring oscillator and of a barotropic Rossby-wave equation are used to demonstrate many of the issues. Results from Kalman-filter estimates, and those from finite interval smoothing are analyzed. In the filter (and prediction) problem, entry of data leads to violation of conservation and other invariant rules. A finite interval smoothing method restores the conservation rules, but uncertainties in all such estimation results remain. Convincing trend and other time-dependent determinations in "reanalysis" -like estimates require a full understanding of both models and observations.