Matthias Morzfeld
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
Research Subjects
Matthias Morzfeld has a total of 52 co-authors. Here are the top co-authors by number of shared publications:
- Brendan Cych
Department of Earth, Ocean and Ecological Sciences, University of Liverpool, Liverpool, UK
4 shared publications - Lisa Tauxe
Scripps Institution of Oceanography University of California San Diego La Jolla CA USA
4 shared publications Kevin K. Lin
Department of Mathematics, University of Arizona, USA
3 shared publications- Oliver R. A. Dunbar
California Institute of Technology Pasadena CA USA
3 shared publications Shay Gilpin
Department of Mathematics, University of Arizona, USA
3 shared publicationsDaniel Hodyss
Marine Meteorology Division
2 shared publications
Latest Publications
- Open Access
- 4 November 2026
- Published by Copernicus GmbH
- Vol. 2026, 1-29
Comment on egusphere-2026-5032
Covariance estimation from a small number of samples is a challenging, but necessary task in ensemble data assimilation (DA) and a fundamental problem in statistics and machine learning. Many covariance estimation methods have been developed in geophysics and in statistics, with little exchange of ideas between the two fields. Covariance estimation in statistics and in ensemble DA rely on fundamentally different assumptions, and we study the efficiency and applicability of both approaches to covariance estimation in systematic numerical experiments. Our numerical experiments are designed to feature a correlation structure in which correlation decays with distance globally, but the rate of decay is location-dependent. In such cases, our findings suggest that methods that make even minimal assumptions on the decay of spatial correlations are more accurate than those that do not do so at all. We also describe how to use a generalized Gaspari-Cohn correlation function to design new covariance estimation methods that capture intricate and spatially varying correlation structures and thereby further reduce covariance estimation errors.
- 4 November 2026
- Published by Copernicus GmbH
- Vol. 2026, 1-29
Open AccessComment on egusphere-2026-5032
Covariance estimation from a small number of samples is a challenging, but necessary task in ensemble data assimilation (DA) and a fundamental problem in statistics and machine learning. Many covariance estimation methods have been developed in geophysics and in statistics, with little exchange of ideas between the two fields. Covariance estimation in statistics and in ensemble DA rely on fundamentally different assumptions, and we study the efficiency and applicability of both approaches to covariance estimation in systematic numerical experiments. Our numerical experiments are designed to feature a correlation structure in which correlation decays with distance globally, but the rate of decay is location-dependent. In such cases, our findings suggest that methods that make even minimal assumptions on the decay of spatial correlations are more accurate than those that do not do so at all. We also describe how to use a generalized Gaspari-Cohn correlation function to design new covariance estimation methods that capture intricate and spatially varying correlation structures and thereby further reduce covariance estimation errors.
- Open Access
- 21 October 2026
- Published by Copernicus GmbH
- Vol. 2026, 1-29
Comment on egusphere-2026-5032
Covariance estimation from a small number of samples is a challenging, but necessary task in ensemble data assimilation (DA) and a fundamental problem in statistics and machine learning. Many covariance estimation methods have been developed in geophysics and in statistics, with little exchange of ideas between the two fields. Covariance estimation in statistics and in ensemble DA rely on fundamentally different assumptions, and we study the efficiency and applicability of both approaches to covariance estimation in systematic numerical experiments. Our numerical experiments are designed to feature a correlation structure in which correlation decays with distance globally, but the rate of decay is location-dependent. In such cases, our findings suggest that methods that make even minimal assumptions on the decay of spatial correlations are more accurate than those that do not do so at all. We also describe how to use a generalized Gaspari-Cohn correlation function to design new covariance estimation methods that capture intricate and spatially varying correlation structures and thereby further reduce covariance estimation errors.
- 21 October 2026
- Published by Copernicus GmbH
- Vol. 2026, 1-29
Open AccessComment on egusphere-2026-5032
Covariance estimation from a small number of samples is a challenging, but necessary task in ensemble data assimilation (DA) and a fundamental problem in statistics and machine learning. Many covariance estimation methods have been developed in geophysics and in statistics, with little exchange of ideas between the two fields. Covariance estimation in statistics and in ensemble DA rely on fundamentally different assumptions, and we study the efficiency and applicability of both approaches to covariance estimation in systematic numerical experiments. Our numerical experiments are designed to feature a correlation structure in which correlation decays with distance globally, but the rate of decay is location-dependent. In such cases, our findings suggest that methods that make even minimal assumptions on the decay of spatial correlations are more accurate than those that do not do so at all. We also describe how to use a generalized Gaspari-Cohn correlation function to design new covariance estimation methods that capture intricate and spatially varying correlation structures and thereby further reduce covariance estimation errors.
- Open Access
- 21 September 2026
- Published by Copernicus GmbH in EGUsphere
- Vol. 2026, 1-29
New and existing covariance estimation techniques for ensemble data assimilation
Covariance estimation from a small number of samples is a challenging, but necessary task in ensemble data assimilation (DA) and a fundamental problem in statistics and machine learning. Many covariance estimation methods have been developed in geophysics and in statistics, with little exchange of ideas between the two fields. Covariance estimation in statistics and in ensemble DA rely on fundamentally different assumptions, and we study the efficiency and applicability of both approaches to covariance estimation in systematic numerical experiments. Our numerical experiments are designed to feature a correlation structure in which correlation decays with distance globally, but the rate of decay is location-dependent. In such cases, our findings suggest that methods that make even minimal assumptions on the decay of spatial correlations are more accurate than those that do not do so at all. We also describe how to use a generalized Gaspari-Cohn correlation function to design new covariance estimation methods that capture intricate and spatially varying correlation structures and thereby further reduce covariance estimation errors.
- 21 September 2026
- Published by Copernicus GmbH in EGUsphere
- Vol. 2026, 1-29
Open AccessNew and existing covariance estimation techniques for ensemble data assimilation
Covariance estimation from a small number of samples is a challenging, but necessary task in ensemble data assimilation (DA) and a fundamental problem in statistics and machine learning. Many covariance estimation methods have been developed in geophysics and in statistics, with little exchange of ideas between the two fields. Covariance estimation in statistics and in ensemble DA rely on fundamentally different assumptions, and we study the efficiency and applicability of both approaches to covariance estimation in systematic numerical experiments. Our numerical experiments are designed to feature a correlation structure in which correlation decays with distance globally, but the rate of decay is location-dependent. In such cases, our findings suggest that methods that make even minimal assumptions on the decay of spatial correlations are more accurate than those that do not do so at all. We also describe how to use a generalized Gaspari-Cohn correlation function to design new covariance estimation methods that capture intricate and spatially varying correlation structures and thereby further reduce covariance estimation errors.
- Open Access
- 1 September 2026
- Published by American Geophysical Union (AGU) in Journal of Advances in Modeling Earth Systems
- Vol. 18 (9)
The Ensemble Kalman Inversion Race
Ensemble Kalman methods were initially developed to solve nonlinear data assimilation problems in oceanography but are now popular in applications far beyond their original use cases. Of particular interest is climate model calibration. As hybrid physics and machine‐learning models advance, the number of parameters and complexity of parameterizations in climate models will continue to grow. To fully realize these advances, we must move from laborious hand‐tuning to calibration‐driven model development in rapid iteration cycles. Thus, robust calibration of these parameters plays an increasingly important role. We focus on learning climate model parameters by minimizing the misfit between modeled and observed climate statistics in an idealized setting. Ensemble Kalman methods are a natural choice for this problem because they are derivative‐free, scalable to high dimensions, and robust to noise caused by statistical observations. Given the many variants of ensemble methods proposed, an important question is: Which ensemble Kalman method should be used for climate model calibration ? To answer this question, we perform systematic numerical experiments to explore the relative computational efficiencies of several ensemble Kalman methods. The numerical experiments involve statistical observations of Lorenz‐type models of increasing complexity, frequently used to represent simplified atmospheric systems, with some featuring neural network parameterizations. For each test problem, several ensemble Kalman methods and a derivative‐based method “race” to reach a specified accuracy, and we measure the computational cost required to achieve the desired accuracy. We investigate how prior information and the parameter or data dimensions define the computational costs of the various methods.
- 1 September 2026
- Published by American Geophysical Union (AGU) in Journal of Advances in Modeling Earth Systems
- Vol. 18 (9)
Open AccessThe Ensemble Kalman Inversion Race
Ensemble Kalman methods were initially developed to solve nonlinear data assimilation problems in oceanography but are now popular in applications far beyond their original use cases. Of particular interest is climate model calibration. As hybrid physics and machine‐learning models advance, the number of parameters and complexity of parameterizations in climate models will continue to grow. To fully realize these advances, we must move from laborious hand‐tuning to calibration‐driven model development in rapid iteration cycles. Thus, robust calibration of these parameters plays an increasingly important role. We focus on learning climate model parameters by minimizing the misfit between modeled and observed climate statistics in an idealized setting. Ensemble Kalman methods are a natural choice for this problem because they are derivative‐free, scalable to high dimensions, and robust to noise caused by statistical observations. Given the many variants of ensemble methods proposed, an important question is: Which ensemble Kalman method should be used for climate model calibration ? To answer this question, we perform systematic numerical experiments to explore the relative computational efficiencies of several ensemble Kalman methods. The numerical experiments involve statistical observations of Lorenz‐type models of increasing complexity, frequently used to represent simplified atmospheric systems, with some featuring neural network parameterizations. For each test problem, several ensemble Kalman methods and a derivative‐based method “race” to reach a specified accuracy, and we measure the computational cost required to achieve the desired accuracy. We investigate how prior information and the parameter or data dimensions define the computational costs of the various methods.
- Open Access
- 7 July 2026
- Published by Elsevier in Proceedings of the Combustion Institute
- Vol. 42
Parasitic combustion model-based data assimilation for predicting wave speeds in rotating detonation engines
- 7 July 2026
- Published by Elsevier in Proceedings of the Combustion Institute
- Vol. 42
Open AccessParasitic combustion model-based data assimilation for predicting wave speeds in rotating detonation engines
- Open Access
- 18 August 2025
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 122 (34)
Reply to Xiao et al.: A wet or dry Martian mid-crust, that is only part of the question
- 18 August 2025
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 122 (34)
Open AccessReply to Xiao et al.: A wet or dry Martian mid-crust, that is only part of the question
- Open Access
- 6 February 2025
- Published by Oxford University Press (OUP) in Geophysical Journal International
- Vol. 241 (1) , 567-579
RamBO: randomized blocky Occam, a practical algorithm for generating blocky models and associated uncertainties
SUMMARY: We present new numerical tools for geophysical inversion and uncertainty quantification (UQ), with an emphasis on blocky (piecewise-constant) layered models that can reproduce sharp contrasts in geophysical or geological properties. The new tools are inspired by an ‘old’ and very successful inversion tool: regularized, nonlinear inversion. We combine Occam’s inversion with total variation regularization and a split Bregman method to obtain an inversion algorithm that we call blocky Occam, because it determines the blockiest model that fits the data adequately. To generate an UQ, we use a modified randomize-then-optimize approach (RTO) and call the resulting algorithm RamBO (randomized blocky Occam), because it essentially amounts to running blocky Occam in a randomized parallel for-loop. Blocky Occam and RamBO inherit computational advantages and stability from the combination of Occam’s inversion, split Bregman and RTO, and, therefore, can be expected to be robustly applicable across geophysics.
- 6 February 2025
- Published by Oxford University Press (OUP) in Geophysical Journal International
- Vol. 241 (1) , 567-579
Open AccessRamBO: randomized blocky Occam, a practical algorithm for generating blocky models and associated uncertainties
SUMMARY: We present new numerical tools for geophysical inversion and uncertainty quantification (UQ), with an emphasis on blocky (piecewise-constant) layered models that can reproduce sharp contrasts in geophysical or geological properties. The new tools are inspired by an ‘old’ and very successful inversion tool: regularized, nonlinear inversion. We combine Occam’s inversion with total variation regularization and a split Bregman method to obtain an inversion algorithm that we call blocky Occam, because it determines the blockiest model that fits the data adequately. To generate an UQ, we use a modified randomize-then-optimize approach (RTO) and call the resulting algorithm RamBO (randomized blocky Occam), because it essentially amounts to running blocky Occam in a randomized parallel for-loop. Blocky Occam and RamBO inherit computational advantages and stability from the combination of Occam’s inversion, split Bregman and RTO, and, therefore, can be expected to be robustly applicable across geophysics.
- Open Access
- 8 November 2024
- Published by Oxford University Press (OUP) in Geophysical Journal International
- Vol. 240 (1) , 747-762
Weather at the core: defining and categorizing geomagnetic excursions and reversals
SUMMARY: Paleomagnetic records provide us with information about the extreme geomagnetic events known as excursions and reversals, but the sparsity of available data limits detailed knowledge of the process and timing. To date there are no agreed on criteria for categorizing such events in terms of severity or longevity. In an analogy to categorizing storms in weather systems, we invoke the magnitude of the global (modified) paleosecular variation index $P_{i_D}$ to define the severity of the magnetic field state, ranging in level from 0 to 3, and defined by instantaneous values of $P_{i_D}$ with level 0 being normal ($P_{i_D}\lt 0.5$) to extreme ($P_{i_D}\ge 15$). We denote the time of entry to an excursional (or reversal) event by when $P_{i_D}$ first exceeds 0.5, and evaluate its duration by the time at which $P_{i_D}$ first returns below its median value, termed the end of event threshold. We categorize each excursional event according to the peak level of $P_{i_D}$ during the entire event, with a range from Category-1 (Cat-1) to Cat-3. We explore an extended numerical dynamo simulation containing more than 1200 events and find that Cat-1 events are the most frequent (72 per cent), but only rarely lead to actual field reversals where the axial dipole, $g_1^0$, has reversed sign at the end of the event. Cat-2 account for about 20 per cent of events, with 34 per cent of those leading to actual reversals, while Cat-3 events arise about 8 per cent of the time but are more likely to produce reversals (43 per cent). Higher category events take as much as 10 times longer than Cat-1 events. Two paleomagnetic field models separately cover the Laschamp excursion and Matuyama–Brunhes (M-B) reversal which are Cat-2 events with respective durations of 3.6 and 27.4 kyr. It seems likely that Cat-2 may be an underestimate for M-B due to limitations in the paleomagnetic records. Our overall results suggest no distinction between excursions and reversals other than a reversal having the ending polarity state opposite to that at the start.- 8 November 2024
- Published by Oxford University Press (OUP) in Geophysical Journal International
- Vol. 240 (1) , 747-762
Open AccessWeather at the core: defining and categorizing geomagnetic excursions and reversals
SUMMARY: Paleomagnetic records provide us with information about the extreme geomagnetic events known as excursions and reversals, but the sparsity of available data limits detailed knowledge of the process and timing. To date there are no agreed on criteria for categorizing such events in terms of severity or longevity. In an analogy to categorizing storms in weather systems, we invoke the magnitude of the global (modified) paleosecular variation index $P_{i_D}$ to define the severity of the magnetic field state, ranging in level from 0 to 3, and defined by instantaneous values of $P_{i_D}$ with level 0 being normal ($P_{i_D}\lt 0.5$) to extreme ($P_{i_D}\ge 15$). We denote the time of entry to an excursional (or reversal) event by when $P_{i_D}$ first exceeds 0.5, and evaluate its duration by the time at which $P_{i_D}$ first returns below its median value, termed the end of event threshold. We categorize each excursional event according to the peak level of $P_{i_D}$ during the entire event, with a range from Category-1 (Cat-1) to Cat-3. We explore an extended numerical dynamo simulation containing more than 1200 events and find that Cat-1 events are the most frequent (72 per cent), but only rarely lead to actual field reversals where the axial dipole, $g_1^0$, has reversed sign at the end of the event. Cat-2 account for about 20 per cent of events, with 34 per cent of those leading to actual reversals, while Cat-3 events arise about 8 per cent of the time but are more likely to produce reversals (43 per cent). Higher category events take as much as 10 times longer than Cat-1 events. Two paleomagnetic field models separately cover the Laschamp excursion and Matuyama–Brunhes (M-B) reversal which are Cat-2 events with respective durations of 3.6 and 27.4 kyr. It seems likely that Cat-2 may be an underestimate for M-B due to limitations in the paleomagnetic records. Our overall results suggest no distinction between excursions and reversals other than a reversal having the ending polarity state opposite to that at the start. - Open Access
- 30 August 2024
- Published by American Geophysical Union (AGU) in Journal of Advances in Modeling Earth Systems
- Vol. 16 (9)
High‐Dimensional Covariance Estimation From a Small Number of Samples
We synthesize knowledge from numerical weather prediction, inverse theory, and statistics to address the problem of estimating a high‐dimensional covariance matrix from a small number of samples. This problem is fundamental in statistics, machine learning/artificial intelligence, and in modern Earth science. We create several new adaptive methods for high‐dimensional covariance estimation, but one method, which we call Noise‐Informed Covariance Estimation (NICE), stands out because it has three important properties: (a) NICE is conceptually simple and computationally efficient; (b) NICE guarantees symmetric positive semi‐definite covariance estimates; and (c) NICE is largely tuning‐free. We illustrate the use of NICE on a large set of Earth science–inspired numerical examples, including cycling data assimilation, inversion of geophysical field data, and training of feed‐forward neural networks with time‐averaged data from a chaotic dynamical system. Our theory, heuristics and numerical tests suggest that NICE may indeed be a viable option for high‐dimensional covariance estimation in many Earth science problems.
- 30 August 2024
- Published by American Geophysical Union (AGU) in Journal of Advances in Modeling Earth Systems
- Vol. 16 (9)
Open AccessHigh‐Dimensional Covariance Estimation From a Small Number of Samples
We synthesize knowledge from numerical weather prediction, inverse theory, and statistics to address the problem of estimating a high‐dimensional covariance matrix from a small number of samples. This problem is fundamental in statistics, machine learning/artificial intelligence, and in modern Earth science. We create several new adaptive methods for high‐dimensional covariance estimation, but one method, which we call Noise‐Informed Covariance Estimation (NICE), stands out because it has three important properties: (a) NICE is conceptually simple and computationally efficient; (b) NICE guarantees symmetric positive semi‐definite covariance estimates; and (c) NICE is largely tuning‐free. We illustrate the use of NICE on a large set of Earth science–inspired numerical examples, including cycling data assimilation, inversion of geophysical field data, and training of feed‐forward neural networks with time‐averaged data from a chaotic dynamical system. Our theory, heuristics and numerical tests suggest that NICE may indeed be a viable option for high‐dimensional covariance estimation in many Earth science problems.
- Open Access
- 12 August 2024
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 121 (35)
Liquid water in the Martian mid-crust
Large volumes of liquid water transiently existed on the surface of Mars more than 3 billion years ago. Much of this water is hypothesized to have been sequestered in the subsurface or lost to space. We use rock physics models and Bayesian inversion to identify combinations of lithology, liquid water saturation, porosity, and pore shape consistent with the constrained mid-crust (∼11.5 to 20 km depths) seismic velocities and gravity near the InSight lander. A mid-crust composed of fractured igneous rocks saturated with liquid water best explains the existing data. Our results have implications for understanding Mars’ water cycle, determining the fates of past surface water, searching for past or extant life, and assessing in situ resource utilization for future missions.
- 12 August 2024
- Published by National Academy of Sciences in Proceedings of the National Academy of Sciences
- Vol. 121 (35)
Open AccessLiquid water in the Martian mid-crust
Large volumes of liquid water transiently existed on the surface of Mars more than 3 billion years ago. Much of this water is hypothesized to have been sequestered in the subsurface or lost to space. We use rock physics models and Bayesian inversion to identify combinations of lithology, liquid water saturation, porosity, and pore shape consistent with the constrained mid-crust (∼11.5 to 20 km depths) seismic velocities and gravity near the InSight lander. A mid-crust composed of fractured igneous rocks saturated with liquid water best explains the existing data. Our results have implications for understanding Mars’ water cycle, determining the fates of past surface water, searching for past or extant life, and assessing in situ resource utilization for future missions.