Search NASA⌕ Search

SEARCH · Search NASA

Results for “SCALE MODELS”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 991 records · Page 55

Simulating Biomass Fast Pyrolysis at the Single Particle Scale

Simulating fast pyrolysis at the scale of single particles allows for the investigation of the impacts of feedstock-specific parameters such as particle size, shape, and species of origin. For this reason particle-scale modeling has emerged as an important tool for understanding how variations in feedstock properties affect the outcomes of pyrolysis processes. The origins of feedstock properties are largely dictated by the composition and hierarchical structure of biomass, from the microstructural porosity to the external morphology of milled particles. These properties may be accounted for in simulations of fast pyrolysis by several different computational approaches depending on the level of structural and chemical complexity included in the model. The predictive utility of particle-scale simulations of fast pyrolysis can still be enhanced substantially by advancements in several areas. Most notably, considerable progress would be facilitated by the development of pyrolysis kinetic schemes that are decoupled from transport phenomena, predict product evolution from whole-biomass with increased chemical speciation, and are still tractable with present-day computational resources.

Ciesielski, Peter↗

Acoustically Induced Streaming Flows near a Model Cod Otolith and their Potential Implications for Fish Hearing

The ears of fishes are remarkable sensors for the small acoustic disturbances associated with underwater sound. For example, each ear of the Atlantic cod (Gadus morhua) has three dense bony bodies (otoliths) surrounded by fluid and tissue, and detects sounds at frequencies from 30 to 500 Hz. Atlantic cod have also been shown to localize sounds. However, how their ears perform these functions is not fully understood. Steady streaming, or time-independent, flows near a 350% scale model Atlantic cod otolith immersed in a viscous fluid were studied to determine if these fluid flows contain acoustically relevant information that could be detected by the ear s sensory hair cells. The otolith was oscillated sinusoidally at various orientations at frequencies of 8 24 Hz, corresponding to an actual frequency range of 280 830 Hz. Phaselocked particle pathline visualizations of the resulting flows give velocity, vorticity, and rate of strain fields over a single plane of this mainly two-dimensional flow. Although the streaming flows contain acoustically relevant information, the displacements due to these flows are likely too small to explain Atlantic cod hearing abilities near threshold. The results, however, may suggest a possible mechanism for detection of ultrasound in some fish species.

54 ENVIRONMENTAL SCIENCES↗

Simulations of ion acceleration at non-relativistic shocks. III. Particle diffusion

We use large hybrid (kinetic-protons-fluid-electrons) simulations to investigate the transport of energetic particles in self-consistent electromagnetic configurations of collisionless shocks. In previous papers of this series, we showed that ion acceleration may be very efficient (up to 10%-20% in energy), and outlined how the streaming of energetic particles amplifies the upstream magnetic field. Here, we measure particle diffusion around shocks with different strengths, finding that the mean free path for pitch-angle scattering of energetic ions is comparable with their gyroradii calculated in the self-generated turbulence. For moderately strong shocks, magnetic field amplification proceeds in the quasi-linear regime, and particles diffuse according to the self-generated diffusion coefficient, i.e., the scattering rate depends only on the amount of energy in modes with wavelengths comparable with the particle gyroradius. For very strong shocks, instead, the magnetic field is amplified up to non-linear levels, with most of the energy in modes with wavelengths comparable to the gyroradii of highest-energy ions, and energetic particles experience Bohm-like diffusion in the amplified field. We also show how enhanced diffusion facilitates the return of energetic particles to the shock, thereby determining the maximum energy that can be achieved in a given time via diffusive shock acceleration. The parameterization of the diffusion coefficient that we derive can be used to introduce self-consistent microphysics into large-scale models of cosmic ray acceleration in astrophysical sources, such as supernova remnants and clusters of galaxies.

79 ASTRONOMY AND ASTROPHYSICS↗

Test Program for High Efficiency Gas Turbine Exhaust Diffuser

This research relates to improving the efficiency of flow in a turbine exhaust, and thus, that of the turbine and power plant. The Phase I SBIR project demonstrated the technical viability of “strutlets” to control stalls on a model diffuser strut. Strutlets are a novel flow-improving vane concept intended to improve the efficiency of flow in turbine exhausts. Strutlets can help reduce turbine back pressure, and incrementally improve turbine efficiency, increase power, and reduce greenhouse gas emmission. The long-term goal is a 0.5 percent improvement of each item, averaged over the US gas turbine fleet. The strutlets were tested in a physical scale model of a gas turbine exhaust diffuser. The test flow passage is a straight, annular diffuser with three sets of struts. At the end of Phase 1, the ability of strutlets to keep flow attached to struts was demonstrated, but the strutlet drag was too high for a net efficiency advantage. An independently sponsored followup project did develop a highly-modified low-drag strutlet. In combination with other flow improving vanes, complicance to the stated goals was demonstrated for for simple cycle power plants, and to most of the goals for combined cycle power plants using this particular exhaust geometry. Importantly, low frequency diffuser noise was reduced by 5 dB or more, compared to the baseline. Appolicability to other diffuser geometries is yet to be demonstrated.

15 GEOTHERMAL ENERGY↗

BioEarth: Envisioning and developing a new regional earth system model to inform natural and agricultural resource management

As managers of agricultural and natural resources are confronted with uncertainties in global change impacts, the complexities associated with the interconnected cycling of nitrogen, carbon, and water present daunting management challenges. Existing models provide detailed information on specific sub-systems (land, air, water, economics, etc). An increasing awareness of the unintended consequences of management decisions resulting from interconnectedness of these sub-systems, however, necessitates coupled regional earth system models (EaSMs). Decision makers’ needs and priorities can be integrated into the model design and development processes to enhance decision-making relevance and "usability" of EaSMs. BioEarth is a current research initiative with a focus on the U.S. Pacific Northwest region that explores the coupling of multiple stand-alone EaSMs to generate usable information for resource decision-making. Direct engagement between model developers and non-academic stakeholders involved in resource and environmental management decisions throughout the model development process is a critical component of this effort. BioEarth utilizes a "bottom-up" approach, upscaling a catchment-scale model to basin and regional scales, as opposed to the "top-down" approach of downscaling global models utilized by most other EaSM efforts. This paper describes the BioEarth initiative and highlights opportunities and challenges associated with coupling multiple stand-alone models to generate usable information for agricultural and natural resource decision-making.

agricultural↗

Microstructure Scale Lithium-Ion Battery Modeling: Part I. On Through-Plane Heterogeneity, Impact of Mesh Representation, and Differences between Macro- and Microscale Models

Li-ion battery performance and degradation are strongly correlated with the electrode microstructures and can be modeled at different scales, each with their own limitations. Herein, we compare predictions achieved with a macro- and a micro-scale model, that is, respectively, neglecting or considering the microstructural heterogeneity of the composite electrodes, on virtual numerically generated and real microstructures. While both models are in relative agreement at the low charge rates, differences arise for fast charging scenarios and especially for the real, highly heterogenous, microstructures. The microscale model predicts that electrolyte concentration saturation and depletion, respectively, at the back of the cathode and of the anode are exacerbated, and that lithium plating occurs earlier for real microstructures. The present work also indicates that the mesh representation significantly impacts the microscale model predictions, and consequently that microscale models should add surface area as a parameter to consider explicitly surface roughness. This article is the first of a series, with subsequent entries further investigating in-plane heterogeneities, lithium plating, and the impact of microstructure representativity on model predictions.

25 ENERGY STORAGE↗

Magnetic field amplification in nonlinear diffusive shock acceleration including resonant and non-resonant cosmic-ray driven instabilities

We present a nonlinear Monte Carlo model of efficient diffusive shock acceleration where the magnetic turbulence responsible for particle diffusion is calculated self-consistently from the resonant cosmic-ray (CR) streaming instability, together with non-resonant short- and long-wavelength CR-current-driven instabilities. We include the backpressure from CRs interacting with the strongly amplified magnetic turbulence which decelerates and heats the super-Alfvénic flow in the extended shock precursor. Uniquely, in our plane-parallel, steady-state, multi-scale model, the full range of particles, from thermal (∼eV) injected at the viscous subshock to the escape of the highest energy CRs (∼PeV) from the shock precursor, are calculated consistently with the shock structure, precursor heating, magnetic field amplification, and scattering center drift relative to the background plasma. In addition, we show how the cascade of turbulence to shorter wavelengths influences the total shock compression, the downstream proton temperature, the magnetic fluctuation spectra, and accelerated particle spectra. A parameter survey is included where we vary shock parameters, the mode of magnetic turbulence generation, and turbulence cascading. From our survey results, we obtain scaling relations for the maximum particle momentum and amplified magnetic field as functions of shock speed, ambient density, and shock size.

79 ASTRONOMY AND ASTROPHYSICS↗

Stochastic Simulation of Complex Fluid Flows (Progress Report for period 07/01/2016 - 06/30/2018)

At a microscopic scale, fluids are composed of molecules whose positions and velocities are random. This gives rise to thermal fluctuations that span the whole range of scales from the microscopic through the mesoscopic, and even the macroscopic. The inclusion of thermal fluctuations is crucial in multi-scale models, which are an important theme in the research program of the DOE Office of Science, and in particular the ASCR Applied Mathematics program's priority focus area on modeling of complex systems involving processes that span vastly different time and/or length scales. In this five-year Early Career project, the PI Aleksandar Donev and collaborators developed computational algorithms for modeling complex fluid mixtures at small scales using a formulation based on fluctuating hydrodynamics. Novel computational methods were developed to model complex fluids with increasing physical complexity, starting from binary miscible and immiscible mixtures, going through multispecies non-reactive and reactive mixtures, and culminating with reactive electrolytes mixtures of neutral molecules and ions. In close collaboration with the group of John Bell at Lawrence Berkeley National Laboratory, the methods were implemented in a scalable computational framework suitable for modern parallel supercomputers, and made publicly available on github. A number of physical examples in which giant nonequilibrium fluctuations are improtant were studied, with a special focus on instabilities at a liquid-liquid interface driven by gravity, diffusion, reactions, and/or electric fields. The methods and codes developed in this project are expected to enable other novel applications in the DOE Basic Energy Sciences program, and engineering sciences more broadly.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Multiscale Model for Simulating Fission-Induced Recrystallization in U-Mo Alloy

U-Mo alloy is one of the most promising fuels for the future high performance research and test reactors due to its high uranium loading density and stable irradiation behavior. However, a major concern to its qualification is the fuel swelling at high fission density due to the formation of large fission-gas bubbles. The fission-induced recrystallization further expedites the formation and growth of gas bubbles. So far, the exact mechanism of recrystallization is still unclear. One possible driving force for recrystallization is the production of point defects and interstitial loops due to irradiation. Following this, we developed a multi-scale model by integrating the rate-theory and phase-field models to study microstructural evolutions in irradiated U-Mo alloy, i.e., fission-induced gas bubble evolution and recrystallization. Density function theory (DFT) calculations are utilized to predict the defect formation energies, diffusivities of defects, the interfacial energy of gas bubble, and the elastic constants of U-Mo alloy, etc. The dislocation density evolution is predicted by the rate-theory developed by Jeff. The grain subdivision occurs once the dislocation density is above a critical value. The grain growth coupled with gas bubbles evolution is modeled using a multi-phase phase-field model. The effects of fission rate and initial grain microstructure on the recrystallization process are systematically investigated. The simulated dislocation density, sub-grain size distribution, and volume fraction of the recrystallization grains under different fission density are comparable with experimental data. (authors)

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Physical Properties of Hanford Transuranic Waste

The research described herein was undertaken to provide needed physical property descriptions of the Hanford transuranic tank sludges under conditions that might exist during retrieval, treatment, packaging and transportation for disposal. The work addressed the development of a fundamental understanding of the types of systems represented by these sludge suspensions through correlation of the macroscopic rheological properties with particle interactions occurring at the colloidal scale in the various liquid media. The results of the work have advanced existing understanding of the sedimentation and aggregation properties of complex colloidal suspensions. Bench scale models were investigated with respect to their structural, colloidal and rheological properties that should be useful for the development and optimization of techniques to process the wastes at various DOE sites.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Helium bubble size effects on the surface morphological response of plasma-facing tungsten

Abstract We report a simulation study on the effects of helium (He) bubble size on the morphological evolution and pattern formation on the surface of tungsten used as a plasma-facing component (PFC) in nuclear fusion devices. We have carried out a systematic investigation based on self-consistent dynamical simulations of surface morphological evolution according to an atomistically-informed, 3D continuum-scale model that captures well the relevant length and time scales of surface nanostructure formation in PFC tungsten. The model accounts for PFC surface diffusion, driven by the biaxial compressive stress originating from the over-pressurized He bubbles in the near-surface region of PFC tungsten as a result of He plasma exposure, combined with the formation of self-interstitial atoms in tungsten that diffuse toward the PFC surface and the flux of surface adatoms generated as a result of surface vacancy-adatom pair formation upon He implantation; this transport of surface adatoms contributes to the anisotropic growth of surface nanostructural features due to the different rates of adatom diffusion along and across step edges of islands on the tungsten surface. Our detailed analysis reveals that varying the average He bubble size plays an important role in the PFC surface growth kinetics as well as the resulting surface topography. Specifically, we find that an increase in the He bubble size leads to a deceleration in the growth rate of the tungsten nanotendrils that emanate from the PFC surface. We also find that the separation distance between the resulting surface features increases with increasing He bubble size, as well as over time. This coarsening effect is a thermally activated process resulting in an accurate description of the temperature dependence of the average surface feature separation by an Arrhenius relation.

36 MATERIALS SCIENCE↗

Autonomous hybrid optimization of a SiO 2 plasma etching mechanism

Computational modeling of plasma etching processes at the feature scale relevant to the fabrication of nanometer semiconductor devices is critically dependent on the reaction mechanism representing the physical processes occurring between plasma produced reactant fluxes and the surface, reaction probabilities, yields, rate coefficients, and threshold energies that characterize these processes. The increasing complexity of the structures being fabricated, new materials, and novel gas mixtures increase the complexity of the reaction mechanism used in feature scale models and increase the difficulty in developing the fundamental data required for the mechanism. This challenge is further exacerbated by the fact that acquiring these fundamental data through more complex computational models or experiments is often limited by cost, technical complexity, or inadequate models. In this paper, we discuss a method to automate the selection of fundamental data in a reduced reaction mechanism for feature scale plasma etching of SiO 2 using a fluorocarbon gas mixture by matching predictions of etch profiles to experimental data using a gradient descent (GD)/Nelder–Mead (NM) method hybrid optimization scheme. These methods produce a reaction mechanism that replicates the experimental training data as well as experimental data using related but different etch processes.

36 MATERIALS SCIENCE↗

A Novel Method to Determine the Hydrodynamic Coefficients of an Eyeball ROV

A good dynamics model is essential and critical for the successful design of navigation and control system of an underwater vehicle. However, it is difficult to determine from the hydrodynamic forces, the inertial added mass terms and the drag coefficients. In this paper, a new experimental method has been used to find the hydrodynamic forces for the ROV II, a remotely operated underwater vehicle. The proposed method is based on the classical free decay test, but with the spring oscillation replaced by a pendulum motion. The experiment results determined from the free decay test of a scaled model compared well with the simulation results obtained from well‐established computational fluid dynamics (CFD) program. Thus, the proposed approach can be used to find the added mass and drag coefficients for other underwater vehicles.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

The birth of a field through a marriage of scales: urban meteorological modeling meets regional climate modeling

Urban meteorological modeling and regional climate modeling developed largely independently, with each discipline addressing different aspects of atmospheric processes across space and time. In this review, I show that urban climate modeling did not arise as a simple scaling extension of urban meteorological modeling or an add-on to regional climate modeling, but instead emerged through the selective inheritance of complementary strengths from its parents after both disciplines reached sufficient methodological maturity. By tracing the parallel evolution of these disciplines, I demonstrate that urban climate modeling inherited the physically explicit treatment of the built environment developed within urban meteorological modeling, and the hierarchical scale translation and climatological framing that matured within regional climate modeling, allowing urban effects to influence climate-relevant outcomes. Building on this synthesis, I propose an objective, and methodologically-grounded definition of urban climate modeling that distinguishes it from urban meteorological modeling. This distinction is increasingly important as urban climate data informs decisions with long-lived societal consequences, and as ambiguity in terminology risks conflating fundamentally different modeling frameworks with distinct physical meaning and decision relevance. In addition to a clarifying definition, this review outlines a research framework for advancing urban climate modeling through scale-aware coupling strategies that preserve physically consistent urban-atmosphere interactions.

regional climate↗

Scaling solid resuspension and sorption for the small column ion exchange processing tank

The Small Column Ion Exchange (SCIX) process is being developed to remove cesium, strontium, and actinides from Savannah River Site (SRS) Liquid Waste using an existing 1.3 million gallon waste tank (i.e., Tank 41H) to house the process. Savannah River National Laboratory (SRNL) is conducting pilot-scale mixing tests to determine the pump requirements for suspending and resuspending Monosodium Titanate (MST), Crystalline Silicotitanate (CST), and simulated sludge. In addition, SRNL will also be conducting pilot-scale tests to determine the mixing requirements for the strontium and actinide sorption. As part of this task, the results from the pilot-scale tests must be scaled up to a full-scale waste tank. This document describes the scaling approach. The pilot-scale tank is a 1/10.85 linear scale model of Tank 41H. The tank diameter, tank liquid level, pump nozzle diameter, pump elevation, and cooling coil diameter are all 1/10.85 of their dimensions in Tank 41H. The pump locations correspond to the proposed locations in Tank 41H by the SCIX Program (Risers B5 and B2 for two pump configurations and Risers B5, B3, and B1 for three pump configurations). MST additions are through Riser E1, the proposed MST addition riser in Tank 41H. To determine the approach to scaling the results from the pilot-scale tank to Tank 41H, the authors took the following approach. They reviewed the technical literature for methods to scale mixing with jets and suspension of solid particles with jets, and the technical literature on mass transfer from a liquid to a solid particle to develop approaches to scaling the test data. SRNL assembled a team of internal experts to review the scaling approach and to identify alternative approaches that should be considered.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Helium segregation on surfaces of plasma-exposed tungsten

Here we report a hierarchical multi-scale modeling study of implanted helium segregation on surfaces of tungsten, considered as a plasma facing component in nuclear fusion reactors. We employ a hierarchy of atomic-scale simulations based on a reliable interatomic interaction potential, including molecular-statics simulations to understand the origin of helium surface segregation, targeted molecular-dynamics (MD) simulations of near-surface cluster reactions, and large-scale MD simulations of implanted helium evolution in plasma-exposed tungsten. We find that small, mobile He-n (1 <= n <= 7) clusters in the near-surface region are attracted to the surface due to an elastic interaction force that provides the thermodynamic driving force for surface segregation. Elastic interaction force induces drift fluxes of these mobile Hen clusters, which increase substantially as the migrating clusters approach the surface, facilitating helium segregation on the surface. Moreover, the clusters' drift toward the surface enables cluster reactions, most importantly trap mutation, in the near-surface region at rates much higher than in the bulk material. Moreover, these near-surface cluster dynamics have significant effects on the surface morphology, near-surface defect structures, and the amount of helium retained in the material upon plasma exposure. We integrate the findings of such atomic-scale simulations into a properly parameterized and validated spatially dependent, continuum-scale reaction-diffusion cluster dynamics model, capable of predicting implanted helium evolution, surface segregation, and its near-surface effects in tungsten. This cluster-dynamics model sets the stage for development of fully atomistically informed coarse-grained models for computationally efficient simulation predictions of helium surface segregation, as well as helium retention and surface morphological evolution, toward optimal design of plasma facing components.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Melt spreading code assessment, modifications, and application to the EPR core catcher design.

The Evolutionary Power Reactor (EPR) is under consideration by various utilities in the United States to provide base load electrical production, and as a result the design is undergoing a certification review by the U.S. Nuclear Regulatory Commission (NRC). The severe accident design philosophy for this reactor is based upon the fact that the projected power rating results in a narrow margin for in-vessel melt retention by external cooling of the reactor vessel. As a result, the design addresses ex-vessel core melt stabilization using a mitigation strategy that includes: (1) an external core melt retention system to temporarily hold core melt released from the vessel; (2) a layer of 'sacrificial' material that is admixed with the melt while in the core melt retention system; (3) a melt plug in the lower part of the retention system that, when failed, provides a pathway for the mixture to spread to a large core spreading chamber; and finally, (4) cooling and stabilization of the spread melt by controlled top and bottom flooding. The overall concept is illustrated in Figure 1.1. The melt spreading process relies heavily on inertial flow of a low-viscosity admixed melt to a segmented spreading chamber, and assumes that the melt mass will be distributed to a uniform height in the chamber. The spreading phenomenon thus needs to be modeled properly in order to adequately assess the EPR design. The MELTSPREAD code, developed at Argonne National Laboratory, can model segmented, and both uniform and nonuniform spreading. The NRC is thus utilizing MELTSPREAD to evaluate melt spreading in the EPR design. MELTSPREAD was originally developed to support resolution of the Mark I containment shell vulnerability issue. Following closure of this issue, development of MELTSPREAD ceased in the early 1990's, at which time the melt spreading database upon which the code had been validated was rather limited. In particular, the database that was utilized for initial validation consisted of: (1) comparison to an analytical solution for the dam break problem, (2) water spreading tests in a 1/10 linear scale model of the Mark I containment by Theofanous et al., and (3) steel spreading tests by Suzuki et al. that were also conducted in a geometry similar to the Mark I. The objective of this work was to utilize the MELTSPREAD code to check the assumption of uniform melt spreading in the EPR core catcher design. As a starting point for the project, the code was validated against the worldwide melt spreading database that emerged after the code was originally written in the very early 1990's. As part of this exercise, the code was extensively modified and upgraded to incorporate findings from these various analytical and experiment programs. In terms of expanding the ability of the code to analyze various melt simulant experiments, the options to input user-specified melt and/or substrate material properties was added. The ability to perform invisicid and/or adiabatic spreading analysis was also added so that comparisons with analytical solutions and isothermal spreading tests could be carried out. In terms of refining the capability to carry out reactor material melt spreading analyses, the code was upgraded with a new melt viscosity model; the capability was added to treat situations in which solid fraction buildup between the liquidus-solidus is non-linear; and finally, the ability to treat an interfacial heat transfer resistance between the melt and substrate was incorporated. This last set of changes substantially improved the predictive capability of the code in terms of addressing reactor material melt spreading tests. Aside from improvements and upgrades, a method was developed to fit the model to the various melt spreading tests in a manner that allowed uncertainties in the model predictions to be statistically characterized. With these results, a sensitivity study was performed to investigate the assumption of uniform spreading in the EPR core catcher that addressed parametric variations in: (1) melt pour mass, (2) melt composition, (3) melt pour rate, (4) pour configuration (i.e., homogeneous vs. stratified metal-oxide phases), (5) melt temperature, (6) cavity condition (wet vs. dry), (7) spreading channel inclination angle, and finally (8) uncertainties in the melt viscosity correlation that are based on comparisons with the reactor material melt spreading database. Although differences were found in the rate of spreading and the degree to which the sacrificial concrete in the spreading room is ablated during the transients, in all cases the melt eventually (over a period of minutes) spreads to a uniform depth in the system.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

Differentiable modelling to unify machine learning and physical models for geosciences

Process-based modelling offers interpretability and physical consistency in many domains of geosciences but struggles to leverage large datasets efficiently. Machine-learning methods, especially deep networks, have strong predictive skills yet are unable to answer specific scientific questions. Here, in this Perspective, we explore differentiable modelling as a pathway to dissolve the perceived barrier between process-based modelling and machine learning in the geosciences and demonstrate its potential with examples from hydrological modelling. ‘Differentiable’ refers to accurately and efficiently calculating gradients with respect to model variables or parameters, enabling the discovery of high-dimensional unknown relationships. Differentiable modelling involves connecting (flexible amounts of) prior physical knowledge to neural networks, pushing the boundary of physics-informed machine learning. It offers better interpretability, generalizability, and extrapolation capabilities than purely data-driven machine learning, achieving a similar level of accuracy while requiring less training data. Additionally, the performance and efficiency of differentiable models scale well with increasing data volumes. Under data-scarce scenarios, differentiable models have outperformed machine-learning models in producing short-term dynamics and decadal-scale trends owing to the imposed physical constraints. Differentiable modelling approaches are primed to enable geoscientists to ask questions, test hypotheses, and discover unrecognized physical relationships. Future work should address computational challenges, reduce uncertainty, and verify the physical significance of outputs.

58 GEOSCIENCES↗