Search NASA⌕ Search

SEARCH · Search NASA

Results for “Reliability Prediction”

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 253 records · Page 14
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

Precision redshift-space galaxy power spectra using Zel'dovich control variates

Numerical simulations in cosmology require trade-offs between volume, resolution and run-time that limit the volume of the Universe that can be simulated, leading to sample variance in predictions of ensemble-average quantities such as the power spectrum or correlation function(s). Sample variance is particularly acute at large scales, which is also where analytic techniques can be highly reliable. This provides an opportunity to combine analytic and numerical techniques in a principled way to improve the dynamic range and reliability of predictions for clustering statistics. In this paper we extend the technique of Zel'dovich control variates, previously demonstrated for 2-point functions in real space, to reduce the sample variance in measurements of 2-point statistics of biased tracers in redshift space. We demonstrate that with this technique, we can reduce the sample variance of these statistics down to their shot-noise limit out to k ~ 0.2 h Mpc -1 . This allows a better matching with perturbative models and improved predictions for the clustering of e.g. quasars, galaxies and neutral Hydrogen measured in spectroscopic redshift surveys at very modest computational expense. We discuss the implementation of ZCV, give some examples and provide forecasts for the efficacy of the method under various conditions.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2025

Benchmarking DFT Accuracy in Predicting O 1s Binding Energies on Metals

X-ray photoelectron spectroscopy (XPS) is a powerful tool for probing the electronic structure and composition of materials, particularly metals and metal oxides of relevance to solar cells and catalysis. Density functional theory (DFT) is often used to support XPS peak assignments, but its reliability for predicting oxygen species is not well established. Here, we compile a large data set of experimental oxygen binding energies and evaluate corresponding DFT predictions. We find that as the binding energies of metal-bound atomic oxygen species increase, especially above ≈530 eV, there is a general decrease in the accuracy of DFTpredicted values. Thus, high-binding-energy atomic oxygen species, such as those proposed as active for selective Ag-catalyzed epoxidation, are less well represented. The chemical nature of the oxygen species also influences accuracy, with molecularly bound species more reliably captured across the entire range of energies. These findings illustrate the limitations of DFT for interpreting XPS spectra and provide a benchmark for improving computational methods.

Adsorption↗
DOE OSTI2025

Harnessing on-machine metrology data for prints with a surrogate model for laser powder directed energy deposition

In this study, we leverage the massive amount of multi-modal on-machine metrology data generated from Laser Powder Directed Energy Deposition (LP-DED) to construct a comprehensive surrogate model of the 3D printing process. By employing Dynamic Mode Decomposition with Control (DMDc), a data-driven technique, we capture the complex physics inherent in this extensive dataset. This physics-based surrogate model emphasizes thermodynamically significant quantities, enabling us to accurately predict key process outcomes. The model ingests 21 process parameters, including laser power, scan rate, and position, while providing outputs such as melt pool temperature, melt pool size, and other essential observables. Furthermore, it incorporates uncertainty quantification to provide bounds on these predictions, enhancing reliability and confidence in the results. We then deploy the surrogate model on a new, unseen part and monitor the printing process as validation of the method. Our experimental results demonstrate that the predictions align with actual measurements with high accuracy, confirming the effectiveness of our approach. Furthermore, this methodology not only facilitates real-time predictions but also operates at process-relevant speeds, establishing a basis for implementing feedback control in LP-DED.

Digital twins↗
DOE OSTI2021

Deciphering Degradation: Machine Learning on Real-World Performance Data (Final Report)

This project addresses a fundamental flaw in solar PV research and solar project financing; the assumed rate of degradation for solar plants. The solar industry currently relies on an out-dated report that observed a 0.5% degradation rate based on a small sample size of systems (~100). While the research conducted at the time was new and innovative, the solar community has not updated this research and universally applies this 0.5% degradation assumption in financial models. Our project updates this assumption by analyzing observed degradation from the industry’s largest dataset of operating solar assets (>10,000 systems) and creating the first machine-learning model based on these observed results to quantify and identify features that drive degradation. There are two strategic goals for this award: reduce the cost of capital (enable solar to attract more capital) and improve the reliability of solar itself. These dual goals are achieved by leveraging an industry dataset to observe system degradation on a large scale, deploying advanced data analysis and machine learning methods to quantify and predict system reliability, and engaging with industry stakeholders to help them accurately price degradation in financial models.

14 SOLAR ENERGY↗
DOE OSTI2023

Bison Verification and Validation Activities for TRISO

Numerical modeling and simulation (M&S) tools play a key role in the research, development, and overall safety assessments of next-generation nuclear energy systems. One such tool, Bison, is a nuclear fuel performance code that is applicable to many fuel forms (e.g., light-water reactor fuel, oxide and metallic fuel for fast reactors, tri-structural isotropic (TRISO) fuel, and plate fuel), and it uses the finite element method to model the thermo- mechanical response of nuclear fuels. One fuel form widely utilized in Generation-IV high-temperature gas-cooled and fluoride- salt-cooled nuclear reactor concepts is TRISO fuel. Recently, Bison’s capabilities were significantly expanded to enable it to model the performance of TRISO particles and compacts. It is important that Bison’s computational results be reliable and predictive, since this code is used to inform high-consequence decisions. The various processes developed to address this issue generally entail two fundamental steps: verification and validation (V&V). Verification ensures that the code functions correctly and is reliable. Code/solution verification, code benchmark, and software quality assurance exercises are examples of verification activities. On the other hand, validation is the process of assessing a code’s capability to accurately model physical problems. Comparisons between code results and experiments quantify the validation level. Application of V&V procedures is crucial to the development of computational tools that are free of coding mistakes and can accurately represent reality. The current study presents an overview of Bison V&V activities relevant to the TRISO fuel concept, which include code/solution verification exercises, CRP-6 Benchmark—a Coordinated Research Program through the International Atomic Energy Agency (IAEA)—exercises, and validation exercises with the Advanced Gas Reactor (AGR)- 1/2/3/4 experiment series.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2026

3D Deep Learning Joint Inversion of Active Seismic Full Waveform and Passive Seismic Traveltime Data for Reservoir Imaging and Uncertainty Quantification

Here, we present deep learning (DL) networks for three-dimensional (3D) joint inversion of active seismic full waveform and passive seismic traveltime data to image reservoirs and their properties and quantify imaging uncertainties. Active seismic full-waveform data can provide high-resolution monitoring images but are collected only intermittently because of their high acquisition cost. In contrast, passive seismic data can be gathered at relatively low cost between regular active surveys, although their imaging quality can be compromised by factors such as low signal-to-noise ratios and limited ray coverage of the target. Although these datasets are routinely acquired together at CO 2 storage sites, their combined inversion within a 3D DL framework has not been previously demonstrated. To our knowledge, this is the first study to address this gap, combining the strength of both data types. For efficient data storage and DL training with large 3D seismic datasets, we use a 3D data matrix in which a random number of passive seismic traveltime data are stored as parabolic envelopes using one-hot encoding and a 3D full-waveform data matrix in which multiple shot gathers are summed. Two network architectures are evaluated: a single-encoder U-Net for single-data type inversion and a dual-encoder U-Net for joint inversion of active and passive seismic data. We also evaluate the single-encoder U-Net for joint inversion by concatenating full-waveform data and traveltime data. We propose a systematic approach for selecting an optimal dropout rate that balances regularization during training and Monte Carlo dropout-based uncertainty quantification during prediction by examining the correlation coefficient between standard deviation and prediction error, along with the training misfit, across a range of dropout rates. 3D DL inversion experiments include five different network configurations, with evaluations under ideal, noisy and dropout-enabled conditions. Both model and data uncertainties are assessed, as well as their combined effects. Across all conditions, the networks consistently predict accurate CO 2 saturation models with low prediction errors, such as a structural similarity index of 0.993 and CO 2 difference of 1.1%. Uncertainty estimates show strong spatial correlation with prediction errors, confirming the effectiveness of the proposed dropout selection approach. The results demonstrate that our DL approach, utilizing compact data representations and appropriate uncertainty quantification, yields accurate subsurface images under various inversion conditions and provides valuable insights into the reliability of predictions.

Um, Evan Schankee [Lawrence Berkeley National Labo↗
DOE OSTI2021

Extracting Material Property Measurement Data from Scientific Articles

Machine learning-based prediction of material properties is often hampered by the lack of sufficiently large training datasets. The majority of such measurement data is embedded in scientific literature and the ability to automatically extract these data is essential to support the development of reliable property prediction methods. In this work, we describe a methodology for an automatic property extraction framework using material solubility as the target property. We create an annotated dataset containing tags for solubility-related entities using a combination of regular expressions and manual tagging. We then compare five entity recognition models leveraging both token-level and span-level architectures on the task of classifying solute names, solubility values, and solubility units. Additionally, we explore a novel pretraining approach that leverages automated chemical name and quantity extraction tools to generate large datasets that do not rely on intensive manual effort. Finally, we perform an analysis to identify the causes of classification errors.

Panapitiya, Gihan U.↗
DOE OSTI2023

Review of machine learning and deep learning models for toxicity prediction

The ever-increasing number of chemicals has raised public concerns due to their adverse effects on human health and the environment. To protect public health and the environment, it is critical to assess the toxicity of these chemicals. Traditional in vitro and in vivo toxicity assays are complicated, costly, and time-consuming and may face ethical issues. These constraints raise the need for alternative methods for assessing the toxicity of chemicals. Recently, due to the advancement of machine learning algorithms and the increase in computational power, many toxicity prediction models have been developed using various machine learning and deep learning algorithms such as support vector machine, random forest, k-nearest neighbors, ensemble learning, and deep neural network. This review summarizes the machine learning- and deep learning-based toxicity prediction models developed in recent years. Support vector machine and random forest are the most popular machine learning algorithms, and hepatotoxicity, cardiotoxicity, and carcinogenicity are the frequently modeled toxicity endpoints in predictive toxicology. It is known that datasets impact model performance. The quality of datasets used in the development of toxicity prediction models using machine learning and deep learning is vital to the performance of the developed models. The different toxicity assignments for the same chemicals among different datasets of the same type of toxicity have been observed, indicating benchmarking datasets is needed for developing reliable toxicity prediction models using machine learning and deep learning algorithms. This review provides insights into current machine learning models in predictive toxicology, which are expected to promote the development and application of toxicity prediction models in the future.

Research & Experimental Medicine↗
DOE OSTI2021

Accomplishments and Year-End Performance Report; Wind Energy Program: Fiscal Year 2021

The National Wind Technology Center (NWTC), located at the U.S. Department of Energy's (DOE's) National Renewable Energy Laboratory (NREL) Flatirons Campus, has been a driving force in advancing wind energy technology research worldwide since its designation as a DOE national research center in 1992. Enabled by the Flatirons Campus's world-class facilities, scientists, engineers, analysts, and researchers are pushing the frontiers of science to pursue wind energy innovation. In Fiscal Year (FY) 2021, NREL continued to provide the technical expertise, research capabilities, and industry understanding to support DOE's ambitious climate action and research goals by advancing technology, addressing market and deployment barriers, and driving down costs with more efficient, reliable, and predictable wind energy systems. One of several highlights, NREL received an R&D 100 Special Recognition Award for its Thermoplastic Resin System for Wind Turbine Blades. This breakthrough in the wind turbine manufacturing process will enable the production of recyclable blades that are stronger, longer, and less expensive, while increasing energy capture, decreasing energy and transportation costs, and increasing blade reliability. In a year when the entire U.S. economy struggled to address workforce gaps, an NREL study compared wind industry needs, training programs, and hiring practices with perspectives from students and recent college graduates. Researchers hope that, by pinpointing areas of disconnect, the expectations of employers who have difficulty filling entry-level jobs can better align with the preparation of the potential applicants who find it hard to break into the field. The lab also made numerous new data and modeling resources available in FY 2021. Recent NREL releases include a modeling tool for predicting the power performance and structural loads of wind turbines within a wind farm (FAST.Farm), a computational framework for modeling golden eagle behavior near wind farms, and 20 years of offshore wind data. Updates were also made to the widely used Wind Plant Integrated Systems Design and Engineering Model (WISDEM), which couples engineering and cost models to examine system-level trade-offs. Now, bolstered by a renewed national commitment to tackle climate change and revitalize the U.S. economy through increased investment in clean energy - particularly in offshore wind energy - NREL stands poised to lead the way to a sustainable future that powers the United States with significant levels of reliable, low-cost, accessible wind energy. This report provides an overview of the achievements NREL made on behalf of DOE's Wind Energy Technologies Office (WETO) and other partners during FY 2021 (between Oct. 1, 2020, and Sept. 30, 2021).

Flatirons Campus↗
DOE OSTI2026

Uncertainty-Aware Machine Learning for Small-Angle X-ray Scattering Analysis in Autonomous Experimentation

Small-angle X-ray scattering (SAXS) is a powerful high-throughput characterization tool for probing nanoscale structure in native sample environments, providing real-time morphological information such as nanoparticle size and shape during synthesis. However, automated SAXS data analysis for extracting meaningful structural parameters is non-trivial and remains a bottleneck in closed-loop experimentation towards autonomous materials discovery, which demands fast, reliable, and uncertainty-aware data analysis. Here, we develop a machine-learning approach for automated SAXS analysis tailored to closed-loop nanoparticle synthesis. A Random Forest (RF) regression model is trained on 100,000 synthetic SAXS curves generated from polydisperse spherical nanoparticles with realistic background contributions. Using normalized one-dimensional SAXS intensity profiles as input, the RF model directly predicts nanoparticle radius, size polydispersity, and background parameters, while the ensemble standard deviation across trees provides built-in uncertainty quantification (UQ). On synthetic data, we show that combining fit-quality metrics (R 2 , MAE) with thresholds on prediction uncertainty reliably identifies accurate parameter estimates without access to ground truth. We then apply the trained model to 365 experimental SAXS profiles of citrate-reduced gold nanoparticles synthesized using an automated droplet-flow microreactor with in situ SAXS at a synchrotron beamline, classifying the results into high- and low-confidence subsets based on UQ metrics. Finally, we integrate RF-based SAXS analysis into a simulated closed-loop optimization campaign using Gaussian process Bayesian optimization to minimize nanoparticle polydispersity, benchmarking against conventional automated Levenberg–Marquardt fitting. The RF-guided campaign exhibits substantially faster convergence and lower relative opportunity cost (∼0.07 vs ∼0.3), demonstrating that uncertainty-aware machine-learning SAXS analysis significantly enhances the efficiency and robustness of autonomous nanomaterials synthesis workflows.

Bayesian optimization↗
DOE OSTI2024

Influence of alloy solidification path on melt pool behavior in additive manufacturing

Numerical models used to study transport phenomena in laser-powder bed fusion processes often rely on assumptions and simplifications to reduce their computational expense. One common simplification is in the description of latent heat evolution during the solid-liquid phase change (i.e., the solidification pathway), justified by the fact that the mushy zone thickness is similar to the numerical grid spacing used for continuum transport models. The lack of resolution of transport phenomena in the mushy zone motivates the use of computationally convenient solidification paths such as linear or sigmoidal relationships over pathways derived from fundamental solidification theory such as equilibrium or Scheil models. In the present work, an uncertainty quantification (UQ) framework is used to analyze the influence of solidification pathway selection on the solidification dynamics and melt pool geometries in laser based additive manufacturing (AM) of IN625. Results show the solidification pathway has a quantifiable influence on the cooling rate at the liquidus isotherm, mushy zone thickness, and solidification time. Due to similarities in the latent heat evolution at the beginning of solidification, the equilibrium and Scheil models predict similar cooling rates near the liquidus isotherm, however the wider freezing range of Scheil leads to a wider mushy zone compared to equilibrium. The non-physical latent heat release profiles of sigmoidal and linear paths lead to significant overpredictions of cooling rates at the liquidus isotherm compared to equilibrium and Scheil. Finally, these results indicate that careful consideration should be given to the choice of solidification pathway to ensure reliable model predictions.

36 MATERIALS SCIENCE↗
DOE OSTI2020

Characterization and verification of a two-body wave energy converter with a novel power take-off

The lack of high efficient, predictable and reliable power take-off (PTO) systems limits developments of ocean wave energy technology. In this paper, a two-body self-reacting wave energy converter (WEC) with a novel PTO is designed, modelled and implemented for efficiency enhancement. A novel mechanical motion rectifier (MMR) using a ball screw mechanism and an enclosed gear set is integrated to improve the energy harvesting efficiency and reliability by rectifying the oscillatory wave motion into unidirectional rotation of the generator. Detailed design and dynamic modelling for the proposed WEC are presented. A prototype of the PTO is tested in a dry lab to characterize and refine the dynamic modelling. The characterized PTO model is combined with the WEC model to create an overall system model. A water tank test is conducted to verify the overall system dynamics, which proves the accuracy of the model and shows the advantages of the proposed WEC on efficiency and predictability. Following the method of how the WEC system is characterized, performance prediction of the proposed WEC with MMR PTO can be achieved with high accuracy.

16 TIDAL AND WAVE POWER↗
DOE OSTI2022

Analysis of Cutter Blade Wear in Rotary Shear Mills

Following development of an analytical abrasive wear model to predict wear of components in a rotary shear mill, the team used a finite element analysis (FEA) approach to calculate forces and loads acting on particles responsible for abrasive wear. The analytical model related critical rotary shear process parameters (shear geometry and rotational speed) to critical material attributes of inorganic mineral species in feedstock (density, size, and aspect ratio) and substrate (hardness and elastic modulus) that enabled us to model the wear of shear cutters in a rotary shear milling system developed by Forest Concepts. With proper knowledge of the forces acting between abrasive particles and cutter components built into it, the model can accurately predict wear of the cutters and provides a quality by design (QbD) approach to predict component reliability based on scientific engineering principles in lieu of trial-and-error approaches. During this reporting period, researchers at Oak Ridge National Laboratory applied an FEA package to simulate the local stresses and forces between an abrasive silica particle and two rotary shear cutters as the distance between the cutters decreases. The output of the FEA was used to provide more accurate projections of the loads applied to the particles in the analytical wear model. Comparison of the FEA force calculations are in good agreement with the loads assumed in the analytical predictions.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Materials Degradation in Extreme Environments: Novel In-situ Measurements of Cracking in Molten Salt

The safety and reliability of next-gen reactors, i.e. molten salt reactors (MSR), depends on materials performance and longevity in extreme environments, yet no in-situ, validated measurement techniques for environmentally assisted cracking (EAC), one potential degradation mechanism, exist for these conditions. This project combined the use of in-situ direct current potential drop (DCPD) crack growth rate determination applied to a horizontal load frame with high temperature fittings and novel sample geometries to enable evaluation in MSR environments and develop a full in-situ SCC measurement capability for extreme environments (high conductivity and elevated T), necessary to predict materials reliability. A successful demonstration of DCPD measurement in molten NaNO 3 /KNO 3 salts at 340 o C was accomplished and validated through post-test fractography imaging. This is the first ever public example, to the authors knowledge, of an in-situ crack growth rate measurement in a molten salt environment.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2020

Effects of Reactive Power on Photovoltaic Inverter Reliability and Lifetime

This paper performs research on predicting Photovoltaic (PV) inverters reliability and lifetime based on thermal cycling. Thermal cycling is considered the most important stressors in an inverter system. In order to achieve this, a detailed electro-thermal model of the PV inverter will be developed along with their controllers capable of providing voltage support through reactive power. An in-house inverter was built, and a PV inverter model was developed to match the physical inverter. The PV inverter electrothermal model was validated for different ambient temperatures to match the in-house inverter hardware. The in-house inverter was placed inside a thermal chamber to emulate different ambient temperatures and their losses and temperature rises within the system were measured. After the validation of the model, a reduced order model of the inverter will be implemented to translate the mission profile of ambient temperature and solar irradiance into load profile of junction temperatures of the switches. The junction temperature data will be used to identify the reliability indices and hence predict the useful lifetime of the inverter system. Along with the model to predict useful lifetime of the system, the impact of reactive power on the overall reliability of the system will be studied. The key observation in this paper shows that lifetime of the inverter decreases as the operating power factor moves away from unity.

14 SOLAR ENERGY↗
DOE OSTI2021

Formation and evolution of protostellar accretion discs – II. From 3D simulation to a simple semi-analytic model of Class 0/I discs

Here we use a 3D radiative non-ideal magnetohydrodynamic simulation to investigate the formation and evolution of a young protostellar disc from a magnetized pre-stellar core. The simulation covers the first ${\sim }10\, {\rm kyr}$ after protostar formation and shows a massive, weakly magnetized disc with radius that initially grows and then saturates at ${\sim }30\, {\rm au}$. The disc is gravitationally unstable with prominent large-amplitude spiral arms. We use our simulation results and a series of physical arguments to construct a predictive and quantitative physical picture of Class 0/I protostellar disc evolution from several aspects, including (i) the angular-momentum redistribution in the disc, self-regulated by gravitational instability to make most of the disc marginally unstable; (ii) the thermal profile of the disc, well-approximated by a balance between radiative cooling and accretion heating; and (iii) the magnetic-field strength and magnetic-braking rate inside the disc, regulated by non-ideal magnetic diffusion. Using these physical insights, we build a simple 1D semi-analytic model of disc evolution. We show that this 1D model, when coupled to a computationally inexpensive simulation for the evolution of the surrounding pseudo-disc, can be used reliably to predict disc evolution in the Class 0/I phase. The predicted long-term evolution of disc size, which saturates at ${\sim }30\, {\rm au}$ and eventually shrinks, is consistent with a recent observational survey of Class 0/I discs. Such hierarchical modelling of disc evolution circumvents the computational difficulty of tracing disc evolution through Class 0/I phase with direct, numerically converged simulations.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2020

Cell degradation quantification—a performance metric-based approach

A safe and reliable battery operation needs effective diagnostic tools. A quantitative failure analysis (FA) to enable cell qualification and quantify its effectiveness for reliable and safe operation of rechargeable Li batteries (RLB) is shown here. The method can identify and quantify potential failure based on the state of charge (SOC) under any operating conditions. A precise and accurate electrochemical analytic diagnosis (eCAD) of 14 rechargeable Li || NMC622 cells of the same build are used as an example. The FA by eCAD can quantitatively decipher good, bad and ugly cells in cycle aging. The cell qualification is based on thermodynamic SOC, not experimental conditions. The method provides a quantitative failure mode and effect analysis (FMEA) to reveal diverse “dead Li” formation that affects the reversibility of the Li anode and charge retention in the cell. This cell qualification method highlights the potential to improve cell quality for safe operation, with strong implications for early fault detection, FA, risk mitigation, state estimation and life prediction for reliable and safe RLB operations

25 ENERGY STORAGE↗
DOE OSTI2025

Analyzing and Exploring Training Recipes for Large-Scale Transformer-Based Weather Prediction

Abstract The rapid rise of deep learning (DL) in numerical weather prediction (NWP) has led to a proliferation of models which forecast atmospheric variables with comparable or superior skill than traditional physics-based NWP. However, among these leading DL models, there is a wide variance in both the training settings and architecture used. Further, the lack of thorough ablation studies makes it hard to discern which components are most critical to success. In this work, we show that it is possible to attain high forecast skill even with relatively off-the-shelf architectures, simple training procedures, and moderate compute budgets. Specifically, we train a minimally modified Swin Transformer V2 (SwinV2) on ERA5 data and find that it attains superior skill in terms of mean-square errors of deterministic forecasts when compared against the European Centre for Medium-Range Weather Forecasts’ Integrated Forecasting System (IFS). Almost all DL–NWP systems share a core set of hyperparameters and design decisions. To aid and expedite future DL–NWP research, we present an in-depth, systematic exploration of different loss functions, model sizes and depths, patch sizes, and multistep training objectives. We also examine the model performance with metrics beyond the typical accuracy (ACC) and RMSE and investigate how the performance scales with model size. Through our open-source code, scoring pipelines, and models, we share our findings on key aspects of the training pipeline. These ablations reduce the necessity for expensive hyperparameter tuning and lower the barrier to entry for future DL–NWP research. Significance Statement This study investigates the potential of using large-scale transformer-based models for weather prediction, showing that it is possible to achieve high forecast accuracy with simpler, off-the-shelf architectures. By training a minimally modified SwinV2 transformer on ERA5 data, we show that the model achieves competitive forecast skill in terms of mean-square error for key variables, outperforming the European Centre for Medium-Range Weather Forecasts’ Integrated Forecasting System (IFS) at all lead times. Our findings suggest that effective training strategies, such as multistep fine-tuning and channel-weighted losses, significantly enhance the model’s performance. However, we also highlight that these improvements come with trade-offs in other areas, such as ensemble spread and high-frequency spatial detail. This work highlights the promise of deep learning in improving weather forecasts, which could lead to better preparedness and response to weather events, ultimately benefiting society by providing more reliable weather predictions.

Willard, Jared D. [Lawrence Berkeley National Labo↗