Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Radiation Reliability”

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 451 records · Page 25
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

UV-Induced Degradation of High-Efficiency Solar Cells with Different Architectures

The ultraviolet-induced degradation (UVID) of solar panels is associated with the deterioration of cell performance and reduced reliability of packaging materials. Here we examine the UV stability of different architectures of high-efficiency solar cells without any encapsulation. Identical UV exposure tests were performed at two different labs using UVA-340 fluorescent lamps under different electrical bias configurations (open- or short-circuit) and irradiated cell surfaces (for bifacial technologies). Cell technologies, including heterojunction (HJ), interdigitated back-contact cells (IBC), passivated emitter rear contact (PERC), passivated emitter rear totally-diffused (PERT), are found to be more susceptible to UVID, leading to significant Isc loss (up to 4%) and P max loss (up to 15%) as compared to the conventional back surface field (Al-BSF) cells after exposure to UV irradiation of 8.92 MJ-m -2 -nm -1 at 340 nm. Additionally, the bifacial cells when irradiated from the backside exhibited greater photocurrent loss compared to the front side exposure, indicating potential sensitivity of rear surface passivation to UV radiation.

crystalline silicon↗
DOE OSTI2024

Development of 211 At production at Crocker Nuclear Laboratory (Final Technical Report)

The focus of this effort was the development of the capability to produce and recover 211 At via the process 209 Bi(α, 2n) 211 At at the cyclotron at the Crocker Nuclear Laboratory at UC Davis. This effort required the design and fabrication of an internal production target, the development of a separation and processing area, and upgrades to the cyclotron to increase its reliability. There was a significant delay due to the pandemic lockdown, but ultimately we were able to produce and recover 211 At at the tens of µCi level using a wet chemistry method. Results of the project were presented at conference, but publication of a refereed article has been delayed by some discrepancies in understanding the calibration of total 211 At produced. We plan to repeat the production using internal funding, but first we want to revise our recovery to use a different technique, developed at the Texas A& M Cyclotron Institute.

07 ISOTOPE AND RADIATION SOURCES↗
DOE OSTI2023

Cross Sections for Coherent Elastic and Inelastic Neutrino-Nucleus Scattering

The prospects of extracting new physics signals in coherent elastic neutrino–nucleus scattering (CEνNS) processes are limited by the precision with which the underlying nuclear structure physics, embedded in the weak nuclear form factor, is known. We present calculations of charge and weak nuclear form factors and CEνNS cross sections on 12C, 16O, 40Ar, 56Fe and 208Pb nuclei. We obtain the proton and neutron densities, and charge and weak form factors by solving Hartree–Fock (HF) equations with a Skyrme (SkE2) nuclear potential. We validate our approach by comparing 208Pb and 40Ar charge form factor predictions with available elastic electron scattering data. Since CEνNS experiments at stopped-pion sources are also well suited to measure inelastic charged–current and neutral–current neutrino–nucleus cross sections, we also present calculations for these processes, incorporating a continuum Random Phase Approximation (CRPA) description on top of the HF–SkE2 picture of the nucleus. Providing both coherent as well as inelastic cross sections in a consistent framework, we aim at obtaining a reliable and detailed comparison of the strength of these processes in the energy region below 100 MeV. Furthermore, we attempt to gauge the level of theoretical uncertainty pertaining to the description of the 40Ar form factor and CEνNS cross sections by comparing relative differences between recent microscopic nuclear theory and widely-used phenomenological form factor predictions. Future precision measurements of CEνNS will potentially help in constraining these nuclear structure details that will in turn improve prospects of extracting new physics.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2021

Smart Methane Emission Detection System Development (Final Report)

Working with the Department of Energy’s National Energy Technology Laboratory, Southwest Research Institute® (SwRI®) developed a system to identify methane leaks reliably, accurately, and autonomously at critical midstream sections of the natural gas distribution network in real- time for the purpose of mitigating methane emissions using Optical Gas Imaging (OGI) cameras. SwRI’s Smart Leak Detection – Methane (SLED/M) adds a high degree of automation to the process of methane leak detection to minimize sources of human error, minimize response time to a leak event, and maximize midstream visibility. Furthermore, SwRI has been working towards integrating Quantitative OGI (QOGI) capabilities into this existing technology. By leveraging Deep Learning, SwRI now has the capability to estimate fugitive emission leak rates quickly and reliably, which allows operators to detect emissions, quantify leak rate, prioritize repairs, and validate the repairs in a single instrument. The next generation QOGI technology leverages the same cameras used in Leak Detection and Repair (LDAR) programs, with improvements in safety and speed for traditional quantification-based repairs, ultimately leading to less overhead cost for the operators. The goals for this research were to develop two types of models with the following goals: 1. Run in real-time on the edge (≥ 12 Hz) 2. Classification: Achieve less than 5% false positive detection 3. Classification: Achieve ≥ 95% methane plume detection rate 4. Regression: achieve ≤ 10 standard cubic feet per hour (scfh) prediction > 70% of the time In order to achieve these results, multiple infrared (IR) and other sensors were investigated in tandem with the midwave IR (MWIR) OGI to provide additional information to train the underlying models. Information on atmospheric conditions including humidity, temperature, pressure, and solar radiation was provided by a weather station. Several machine learning and deep learning architectures and methods, including looking at quantized classification networks and regressions networks, were explored. As further data was collected, curated, and labeled, it allowed for more refined regressive networks to be adequately trained, leading to better insight into the true flow rates being observed. An important valuable deliverable of this research effort was the development of an advanced network which underwent multiple iterations capable of giving a continuous output. The current network has a predicted mean average percentage error (MAPE) of 12.3% just outside our target goal of 10.00%, but an accuracy of 97.78% at ±50 scfh, well within the overall goal for the Department of Energy (DOE) program. Upon closer inspection, it was observed that more than 10% of datapoints contributing to the MAPE predictions were the result of low flow rate predictions and are beyond the sensitivity of instrument measurement as a result of normal operational variation and noise.

03 NATURAL GAS↗
DOE OSTI2026

Robust dynamic operation of high temperature electrolysis solid oxide cells

Here, this study evaluates the durability of Ni/YSZ-supported SOECs under dynamic operating conditions relevant to real-world applications. Systematic tests were conducted to assess cell performance under steam cycling (3 to 75% humidified H 2 ), mode cycling between SOFC and SOEC operation, thermal cycling (150 to 750 °C at OCV, and 600 to 800 °C at 1.3V), and redox cycling (between 50% humidified H 2 and 50% humidified N 2 ). Steam cycling, mode cycling, and thermal cycling at OCV do not significantly accelerate performance degradation. Thermal cycling at 1.3V caused minimal damage within 600 to 800 °C. Full redox cycling (multi-hour oxidation holds) induced cell structural failure, while partial redox cycling (0.5 h holds) was tolerated. Extensive characterization revealed some material evolution, namely Sr and Co secondary phase formation, within the oxygen electrode due to La 0.6 Sr 0.4 Co 0.2 Fe 0.8 O 3-δ instability, especially for steam cycling and mode cycling. Minimal changes were identified within the Ni-YSZ fuel electrodes. These findings provide critical insights into SOEC reliability under dynamic conditions, supporting their application in dynamic or intermittent energy systems.

08 HYDROGEN↗
DOE OSTI2024

Rapid Prototyping for Nanoparticle-Based Photonic Crystal Fiber Sensors

The advent of nanotechnology has motivated a revolution in the development of miniaturized sensors. Such sensors can be used for radiation detection, temperature sensing, radio-frequency sensing, strain sensing, and more. At the nanoscale, integrating the materials of interest into sensing platforms can be a common issue. One promising platform is photonic crystal fibers, which can draw in optically sensitive nanoparticles or have its optical properties changed by specialized nanomaterials. However, testing these sensors at scale is limited by the the need for specialized equipment to integrate these photonic crystal fibers into optical fiber systems. Having a method to enable rapid prototyping of new nanoparticle-based sensors in photonic crystal fibers would open up the field to a wider range of laboratories that could not have initially studied these materials in such a way before. This manuscript discusses the improved processes for cleaving, drawing, and rapidly integrating nanoparticle-based photonic crystal fibers into optical system setups. The method proposed in this manuscript achieved the following innovations: cleaving at a quality needed for nanoparticle integration could be done more reliably (≈100% acceptable cleaving yield versus ≈50% conventionally), nanoparticles could be drawn at scale through photonic crystal fibers in a safe manner (a method to draw multiple photonic crystal fibers at scale versus one fiber at a time), and the new photonic crystal fiber mount was able to be finely adjusted when increasing the optical coupling before inserting it into an optical system (before, expensive fusion splicing was the only other method).

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2020

Bayesian Structural Time Series for Behind-the-Meter Photovoltaic Disaggregation

Distributed photovoltaic (PV) generation often occurs ``behind the meter": a grid operator can only observe the net load, which is the sum of the gross load and distributed PV generation. This lack of observability poses a challenge to system operation at both bulk level and distribution level. The lack of real-time or near-future disaggregated estimates of gross load and PV generation will lead to over scheduling of energy production and regulation reserves, reliability constraints violations, wear and tear of controller devices, and potentially cascading failures of a system. In this paper we propose the use of a Bayesian Structural Time Series (BSTS) model with local solar irradiance measurements to disaggregate the summed PV generation and gross load signals at a downstream measurement site. BSTSs are a highly expressive model class that blends classic time series models with the powerful Bayesian state space estimation framework. Disaggregation is done probabilistically, which automatically quantifies the uncertainties of the estimated PV generation and gross load consumption. Depending on the data availability in real-time, it can be used to disaggragate PV and gross load at customer site, or can be used at the feeder level. In this paper, we focus on solving the problem at feeder level. We compare the performance of a BSTS model as well as a handful of state-of-the-art methods on a Pecan Street AMI dataset, using the National Solar Radiation Database (NSRDB) to estimate local irradiance.

bayesian structural time series↗
DOE OSTI2022

Lyman alpha and Lyman continuum emission of Mg Ⅱ -selected star-forming galaxies

We present observations with the Cosmic Origins Spectrograph onboard the Hubble Space Telescope of seven compact low-mass star-forming galaxies at redshifts, z, in the range 0.3161–0.4276, with various O 3 Mg 2 = [O iii] λ5007/Mg ii λ2796+2803 and Mg 2 = Mg ii λ2796/Mg ii λ2803 emission-line ratios. We aim to study the dependence of leaking Lyman continuum (LyC) emission on the characteristics of Mg ii emission together with the dependencies on other indirect indicators of escaping ionizing radiation. LyC emission with escape fractions f esc (LyC) = 3.1–4.6 percent is detected in four galaxies, whereas only 1σ upper limits of f esc (LyC) in the remaining three galaxies were derived. A strong narrow Lyα emission line with two peaks separated by V sep ~ 298–592 km s -1 was observed in four galaxies with detected LyC emission and very weak Lyα emission is observed in galaxies with LyC non-detections. Our new data confirm the tight anticorrelation between f esc (LyC) and V sep found for previous low-redshift galaxy samples. V sep remains the best indirect indicator of LyC leakage among all considered indicators. It is found that escaping LyC emission is detected predominantly in galaxies with Mg 2 ≳ 1.3. A tendency of an increase of f esc (LyC) with increasing of both the O 3 Mg 2 and Mg 2 is possibly present. However, there is substantial scatter in these relations not allowing their use for reliable prediction of f esc (LyC).

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2026

Development of a 95-Year Solar Dataset for Resource Adequacy Studies

Long-term high-resolution solar data provides enhanced understanding of variability of solar generation and enhances our ability to develop strategies for a resilient and reliable electric grid under high deployment of solar energy. Therefore, it is important to develop long-term synthetic datasets that can provide multiple occurrences of various severe weather scenarios that are expected to test the limits of resource adequacy under scenarios contain various energy generation sources. Examples of such scenarios could be long periods of high temperatures when demand for electricity is high or periods where high winds could lead to a shut-down of transmission lines for long periods of time to ensure fire safety. NREL has developed the first version of such a dataset covering a 95-year period covering 2006-2100 at a 4km hourly resolution. This dataset contains all variables necessary to calculate solar generation. During development of this dataset, we focused on creating unbiased, high-resolution solar irradiance through statistical downscaling methods, using Regional Climate Model (RCM) simulations from the North American Coordinated Regional Climate Downscaling Experiment (NA-CORDEX) as input. The National Solar Radiation Database (NSRDB) containing over 25 years of observations was used to calibrate the statistical downscaling models. This presentation will outline the primary steps in developing this dataset, including (1) regridding RCM data to a common grid at 20-km resolution, (2) correcting RCM biases with NSRDB, (3) applying temporal and spatial downscaling methods to generate high-resolution (4-km, hourly) solar and ancillary data. Additionally, we will present an evaluation of the downscaled data against the NSRDB across various zones in the CONUS. Lastly, we will present a user guide for accessing the datasets.

14 SOLAR ENERGY↗
DOE OSTI2021

Implementation of an Experimental Setup Based on the Three-Omega Method for Thermal Conductivity Measurements of Molten Salts

Molten salt reactors (MSRs) are one of the most promising next-generation nuclear reactor designs which use molten salts as the fuel and coolant. To design, model, predict, license, and operate MSRs, the behavior of the fuel and coolant must be understood as a function of temperature and irradiation, i.e., burnup. Of particular importance and complexity is the reliable and reproducible measurement of thermal conductivity of molten salts at high temperatures (> 500 °C). Thermal conductivity provides information on the rate at which heat passes through a specified material and is dependent on chemical composition. Currently, in the national laboratory complex, several methods for measuring thermal conductivity on molten salt systems are under development, however, to date, consistent data have not been produced. Typically, thermal conductivity measurements of high-temperature melts have been conducted by using the Transient Hot Wire (THW) method where thermal conductivity is determined by the change in temperature over time at a certain distance from a heating wire. Another time-dependent technique is the laser flash method, commonly used for solid samples. Laser flash requires prior measurements of density and heat capacity to calculate the thermal conductivity of the sample, introducing additional measurement errors. Steady state and time-dependent measurements such as THW and laser flash have errors from convection and ambient temperature fluctuations (heat radiation). In order to produce high fidelity thermal conductivity data, a measurement technique needs to be developed that will reduce the effect of convection at high temperatures by measuring in a very fast time window. In addition, to measure on actinide and irradiated salts the sample size should be small. Therefore, this project aims to develop a frequency-domain technique based on the three-omega (3?) method, previously proven to work on solids, nanofluids and low temperature solar salts (25 to 250 °C). If the 3?-device can be designed with materials that can withstand high temperatures (>500 °C) and can be configured to reduce errors due to convection, then it would prove to be a valuable technique for ensuring thermal conductivity on molten salt systems needed for MSR development. For this proposal, the 3? instrument will be designed, assembled, and tested on non-radioactive salts before being used on actinide bearing salts. In addition, a thermal model will be developed to determine the thermal conductivity of the sample from the measured thermal oscillation of the sensor. Because 3? is a direct method that requires a small volume of sample and works in a very fast time window, it will aid in suppressing the interference due to convection which is ideal when working with high temperature liquid samples.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2021

Improved naive Bayesian probability classifier in predictions of nuclear mass

Recently, novel statistical methods such as neural networks and Bayesian learning methods are implemented to describe the nuclear masses. Based on previous studies, an improved naive Bayesian probability (iNBP) classifier is proposed to study the nuclear masses by refining the results of sophisticated nuclear models. In the iNBP method, the prediction for nuclear masses is treated as a classification problem. The residuals are classified into several groups to generate prior and conditional probabilities, and the posterior probabilities are further determined by the Bayesian formula. We choose the expectation with maximum probability as the final prediction. Reliability of the iNBP method is assessed by analyzing the global optimizations and the extrapolating capabilities. Here, the iNBP method exhibits impressive improvements on global descriptions for different mass models. Moreover, the method shows robust extrapolating capabilities. Results demonstrate the iNBP method can be applied to predict the nuclear masses of unknown regions. Considering the local mass relations, the iNBP method can offer considerable fine-tuning of the mass descriptions from nuclear models. The methodology proposed in this paper can also be applied to other model-based extrapolations of nuclear observables.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2022

Quasi-deuteron model at low renormalization group resolution

The quasi-deuteron model introduced by Levinger is used to explain cross sections for knocking out high-momentum protons in photoabsorption on nuclei. This is within a framework we characterize as exhibiting high renormalization group (RG) resolution. Assuming a one-body reaction operator, the nuclear wave function must include two-body short-range correlations (SRCs) with deuteronlike quantum numbers. In Phys. Rev. C 104, 034311 (2021), we showed that SRC physics can be naturally accounted for at low RG resolution. We describe the quasi-deuteron model at low RG resolution and determine the Levinger constant, which is proportional to the ratio of nuclear photoabsorption to that for photodisintegration of a deuteron. We extract the Levinger constant based on the ratio of momentum distributions at high relative momentum. We compute momentum distributions evolved under similarity RG (SRG) transformations where the SRC physics is shifted into the operator as a universal two-body term. The short-range nature of this operator motivates using local-density approximations with uncorrelated wave functions in evaluating nuclear matrix elements, which greatly simplifies the analysis. The operator must be consistently matched to the RG scale and scheme of the interaction for a reliable extraction. We apply SRG transformations to different nucleon-nucleon (NN) interactions and use the deuteron wave functions and Weinberg eigenvalues to determine approximate matching scales. We predict the Levinger constant for several NN interactions and a wide range of nuclei comparing to experimental extractions. The predictions at low RG resolution are in good agreement with experiment when starting with a hard NN interaction and the initial operator. Similar agreement is found using soft NN interactions when the additional two-body operator induced by evolution from hard to soft is included.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2025

Summary of FY25 activities on the project Realistic Reaction Evaluations for Fission Products Off Stability (RREFPOS)

Applications such nonproliferation, post-detonation forensics, spent-fuel assay, reactor burnup and design, as well as astrophysics, rely on the accurate description of the neutron interaction with unstable fission products. Current neutron-induced cross-section descriptions for these nuclei are either non-existent or based on simplified assumptions, leading to unquantified impacts on predicted cross-sections. The goal of the Realistic Reaction Evaluations for Fission Products Off Stability (RREFPOS) project is to develop a reproducible methodology leading to a more predictive and realistic new set of evaluated nuclear data, improving the reliability of applications involving fission products off stability. The current report documents the activities related to the RREFPOS project in its first year, corresponding to the Fiscal Year of 2025 (FY25).

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2024

Toward coherent quantum computation of scattering amplitudes with a measurement-based photonic quantum processor

In recent years, applications of quantum simulation have been developed to study the properties of strongly interacting theories. This has been driven by two factors: on the one hand, needs from theorists to have access to physical observables that are prohibitively difficult to study using classical computing; on the other hand, quantum hardware becoming increasingly reliable and scalable to larger systems. In this work, we discuss the feasibility of using quantum optical simulation for studying scattering observables that are presently inaccessible via lattice QCD and are at the core of the experimental program at Jefferson Laboratory, the future Electron-Ion Collider, and other accelerator facilities. We show that recent progress in measurement-based photonic quantum computing can be leveraged to provide deterministic generation of required exotic gates and implementation in a single photonic quantum processor. Published by the American Physical Society 2024

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2026

Detector alignment for X-ray crystallography using Millepede-II

I describe a method for accurately refining the geometrical parameters of segmented X-ray area detectors on the basis of serial crystallography data, using 'Millepede' – an algorithm created for a very similar problem in high-energy physics. The Millepede method for serial crystallography builds on the approach of Brewster et al. [Acta Cryst. (2018), D74, 877–894], in which the detector parameters are refined simultaneously with the parameters for each individual crystal. This accounts for the mutual dependency between the parameters and thereby avoids the bias and slow convergence problems that have afflicted older approaches in which the deviations between observed and calculated Bragg peak positions were taken directly as the updates for the detector panel positions. The Millepede method uses the special structure of the least-squares normal equations to reduce them to a much smaller form that can be solved very quickly, even compared with the sparse matrix methods used previously. This makes it practical to refine the detector geometry frequently and thereby maintain accurate calibration without specialized alignment campaigns. Tilts of detector panels out of the plane can be reliably refined, as can the overall distance of the detector in the beam direction. With a simulated test case, the new method produced panel shifts within 7% of the correct values with only one iteration, and produced almost exactly correct shifts after a second iteration. A simulated out-of-plane panel rotation was correctly determined to within 0.001°. Applied to experimental data from an X-ray free-electron laser, the method increased the indexable fraction of frames from 30% to 91% in a single iteration, and to 96% after two further iterations. Computing the geometry updates on the basis of 2060 crystals took only 0.819 s on desktop computing hardware, including the time taken to read the required data from disk. The scaling was found to be very close to linear for up to 100 980 sets of crystal parameters, which took only 78.2 s to process under the same conditions. The method has been applied as part of a real-time feedback system at a synchrotron radiation beamline, in which an out-of-plane detector tilt of 0.04° was detected and corrected. Possible further applications are also described here.

Millepede-II↗
DOE OSTI2023

Ground Motion Models (GMMs) Improvements Using Earthquake Simulations on High Performance Computers

A computationally efficient simulation platform was developed that can provide representative synthetic ground motions from crustal earthquakes in the Stable Continental Regions of Central and Eastern US (CEUS), using 3D modeling and high-performance computing. The main objective was to use synthetic ground motion to provide constrains to refinements of exiting ergodic Ground Motion Models (GMMs), for large magnitude earthquakes and near-fault distances, for which these models are less reliable. Physics-based broadband (0-5Hz) ground motion simulations were used to estimate the near-fault ground motion amplitudes and within event and between-event variabilities associated with fault rupture characteristics. As part of a strategy for selecting a reginal velocity model and validation of developed rupture modeling technique, ground motions from the moment magnitude Mw5.0 November 7, 2016, Cushing Oklahoma, and Mw5.8 September 3, 2016, Pawnee Oklahoma earthquakes were simulated. In our simulations we used a 3D regional velocity model that was based on Saikia’s 1D velocity model. Saikia’s model demonstrated better performance in modelling high frequency regional wave propagation for CEUS region. The proposed 3D model includes lateral variations added to the 1D background model using the stochastic scheme of Pitarka and Mellors. Comparisons of the simulations with recordings of both earthquakes demonstrated the reliability of our deterministic simulation approach while emphasizing the importance of including small-scale variability in the regional velocity model needed to reproduce the observed high-frequency wave scattering effects. As part of validation analysis, comparisons with different GMMs for a Mw6.5 earthquake in the CESUS region resulted in a very good match between the simulated and empirical ground motion models. Initial investigations of within-event and between-event ground motion variabilities for Mw6.5 scenario earthquakes on a strike-slip fault, suggest that they are strongly related to spatial slip and slip rate variations, average rupture velocity, rupture area and rupture initiation location. For certain scenarios we found that the ground motion variability observed at near-fault distances (< 5 km) also persists at longer distances. Regardless of the rupture scenario, the simulated ground motion tends to fully saturate at short distances and for all periods. The near-fault saturation has to do with the attenuation of waves propagating along the fault and local rupture radiation pattern that also contribute to stronger ground motion variation at such distances. Analysis of effects of rupture initiation location suggest that the peak ground motion (PGV) and spectral acceleration (SA) can be quite variable due to rupture directivity effects. Such effects are stronger at periods longer than 1s. The effect of the 1D velocity models and surface topography on simulated ground motion were investigated by comparing three component synthetic seismograms computed at selected sites. Effect of surface topography was considered using the ratio between spectral accelerations simulated for two 1D models with flat surface topography and realistic model with surface topography. Overall, the topography slightly amplifies (by ~30%) the ground motion amplitude in the frequency range 1-3Hz. The effect of topography is more visible in the surface and coda waves portion of the seismograms.

58 GEOSCIENCES↗
DOE OSTI2024

Data Imbalance, Uncertainty Quantification, and Transfer Learning in Data‐Driven Parameterizations: Lessons From the Emulation of Gravity Wave Momentum Transport in WACCM

Abstract Neural networks (NNs) are increasingly used for data‐driven subgrid‐scale parameterizations in weather and climate models. While NNs are powerful tools for learning complex non‐linear relationships from data, there are several challenges in using them for parameterizations. Three of these challenges are (a) data imbalance related to learning rare, often large‐amplitude, samples; (b) uncertainty quantification (UQ) of the predictions to provide an accuracy indicator; and (c) generalization to other climates, for example, those with different radiative forcings. Here, we examine the performance of methods for addressing these challenges using NN‐based emulators of the Whole Atmosphere Community Climate Model (WACCM) physics‐based gravity wave (GW) parameterizations as a test case. WACCM has complex, state‐of‐the‐art parameterizations for orography‐, convection‐, and front‐driven GWs. Convection‐ and orography‐driven GWs have significant data imbalance due to the absence of convection or orography in most grid points. We address data imbalance using resampling and/or weighted loss functions, enabling the successful emulation of parameterizations for all three sources. We demonstrate that three UQ methods (Bayesian NNs, variational auto‐encoders, and dropouts) provide ensemble spreads that correspond to accuracy during testing, offering criteria for identifying when an NN gives inaccurate predictions. Finally, we show that the accuracy of these NNs decreases for a warmer climate (4 × CO 2 ). However, their performance is significantly improved by applying transfer learning, for example, re‐training only one layer using ∼1% new data from the warmer climate. The findings of this study offer insights for developing reliable and generalizable data‐driven parameterizations for various processes, including (but not limited to) GWs.

54 ENVIRONMENTAL SCIENCES↗