SEARCH · Engineering Papers

Results for “design source code”

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.

SourceResearch connectionFrom yearThrough year

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

At least 469 records · Page 26

All sources NASA NTRS DOE OSTI

NASA NTRS1980

A translator writing system for microcomputer high-level languages and assemblers

In order to implement high level languages whenever possible, a translator writing system of advanced design was developed. It is intended for routine production use by many programmers working on different projects. As well as a fairly conventional parser generator, it includes a system for the rapid generation of table driven code generators. The parser generator was developed from a prototype version. The translator writing system includes various tools for the management of the source text of a compiler under construction. In addition, it supplies various default source code sections so that its output is always compilable and executable. The system thereby encourages iterative enhancement as a development methodology by ensuring an executable program from the earliest stages of a compiler development project. The translator writing system includes PASCAL/48 compiler, three assemblers, and two compilers for a subset of HAL/S.

Collins, W. R.↗

NASA NTRS1992

Simulation and analysis of support hardware for multiple instruction rollback

Recently, a compiler-assisted approach to multiple instruction retry was developed. In this scheme, a read buffer of size 2N, where N represents the maximum instruction rollback distance, is used to resolve one type of data hazard. This hardware support helps to reduce code growth, compilation time, and some of the performance impacts associated with hazard resolution. The 2N read buffer size requirement of the compiler-assisted approach is worst case, assuring data redundancy for all data required but also providing some unnecessary redundancy. By adding extra bits in the operand field for source 1 and source 2 it becomes possible to design the read buffer to save only those values required, thus reducing the read buffer size requirement. This study measures the effect on performance of a DECstation 3100 running 10 application programs using 6 read buffer configurations at varying read buffer sizes.

Alewine, Neil J.↗

NASA NTRS1993

Upwind MacCormack Euler solver with non-equilibrium chemistry

A computer code, designated UMPIRE, is currently under development to solve the Euler equations in two dimensions with non-equilibrium chemistry. UMPIRE employs an explicit MacCormack algorithm with dissipation introduced via Roe's flux-difference split upwind method. The code also has the capability to employ a point-implicit methodology for flows where stiffness is introduced through the chemical source term. A technique consisting of diagonal sweeps across the computational domain from each corner is presented, which is used to reduce storage and execution requirements. Results depicting one dimensional shock tube flow for both calorically perfect gas and thermally perfect, dissociating nitrogen are presented to verify current capabilities of the program. Also, computational results from a chemical reactor vessel with no fluid dynamic effects are presented to check the chemistry capability and to verify the point implicit strategy.

Sherer, Scott E.↗

NASA NTRS1994

Design and Implementation of a Distributed Version of the NASA Engine Performance Program

Distributed NEPP is a new version of the NASA Engine Performance Program that runs in parallel on a collection of Unix workstations connected through a network. The program is fault-tolerant, efficient, and shows significant speed-up in a multi-user, heterogeneous environment. This report describes the issues involved in designing distributed NEPP, the algorithms the program uses, and the performance distributed NEPP achieves. It develops an analytical model to predict and measure the performance of the simple distribution, multiple distribution, and fault-tolerant distribution algorithms that distributed NEPP incorporates. Finally, the appendices explain how to use distributed NEPP and document the organization of the program's source code.

Cours, Jeffrey T.↗

NASA NTRS2007

General Mission Analysis Tool (GMAT)

The General Mission Analysis Tool (GMAT) is a space trajectory optimization and mission analysis system developed by NASA and private industry in the spirit of the NASA Mission. GMAT contains new technology and is a testbed for future technology development. The goal of the GMAT project is to develop new space trajectory optimization and mission design technology by working inclusively with ordinary people, universities, businesses, and other government organizations, and to share that technology in an open and unhindered way. GMAT is a free and open source software system licensed under the NASA Open Source Agreement: free for anyone to use in development of new mission concepts or to improve current missions, freely available in source code form for enhancement or further technology development.

Hughes, Steven P.↗

NASA NTRS2005

Coherent Laser Instrument Would Measure Range and Velocity

A proposed instrument would project a narrow laser beam that would be frequency-modulated with a pseudorandom noise (PN) code for simultaneous measurement of range and velocity along the beam. The instrument performs these functions in a low mass, power, and volume package using a novel combination of established techniques. Originally intended as a low resource- footprint guidance sensor for descent and landing of small spacecraft onto Mars or small bodies (e.g., asteroids), the basic instrument concept also lends itself well to a similar application guiding aircraft (especially, small unmanned aircraft), and to such other applications as ranging of topographical features and measuring velocities of airborne light-scattering particles as wind indicators. Several key features of the instrument s design contribute to its favorable performance and resource-consumption characteristics. A laser beam is intrinsically much narrower (for the same exit aperture telescope or antenna) than a radar beam, eliminating the need to correct for the effect of sloping terrain over the beam width, as is the case with radar. Furthermore, the use of continuous-wave (CW), erbium-doped fiber lasers with excellent spectral purity (narrow line width) permits greater velocity resolution, while reducing the laser s power requirement compared to a more typical pulsed solid-state laser. The use of CW also takes proper advantage of the increased sensitivity of coherent detection, necessary in the first place for direct measurement of velocity using the Doppler effect. However, measuring range with a CW beam requires modulation to "tag" portions of it for time-of-flight determination; typically, the modulation consists of a PN code. A novel element of the instrument s design is the use of frequency modulation (FM) to accomplish both the PN-modulation and the Doppler-bias frequency shift necessary for signed velocity measurements. This permits the use of a single low-power waveguide electrooptic phase modulator, while simultaneously mitigating the effects of speckle as a noise source in the coherent detection.

Chang, Daniel↗

NASA NTRS2015

Big Software for SmallSats: Adapting CFS to CubeSat Missions

Expanding capabilities and mission objectives for SmallSats and CubeSats is driving the need for reliable, reusable, and robust flight software. While missions are becoming more complicated and the scientific goals more ambitious, the level of acceptable risk has decreased. Design challenges are further compounded by budget and schedule constraints that have not kept pace. NASA's Core Flight Software System (cFS) is an open source solution which enables teams to build flagship satellite level flight software within a CubeSat schedule and budget. NASA originally developed cFS to reduce mission and schedule risk for flagship satellite missions by increasing code reuse and reliability. The Lunar Reconnaissance Orbiter, which launched in 2009, was the first of a growing list of Class B rated missions to use cFS. Large parts of cFS are now open source, which has spurred adoption outside of NASA. This paper reports on the experiences of two teams using cFS for current CubeSat missions. The performance overheads of cFS are quantified, and the reusability of code between missions is discussed. The analysis shows that cFS is well suited to use on CubeSats and demonstrates the portability and modularity of cFS code.

Open Source↗

NASA NTRS2015

Big Software for SmallSats: Adapting cFS to CubeSat Missions

Flight Software↗

NASA NTRS2017

Hemispherical Optical Dome for Underwater Communication

For many years, acoustic systems have been used as the primary method for underwater communication; however, the data transfer rate of such systems is low because sound propagates slowly through the water. A higher throughput can be achieved using visible light to transmit data underwater. The first issue with using this approach is that there is generally a large loss of the light signal due to scattering and absorption in water even though there is an optimal wavelength for transmission in the blue or green wavelengths of the visible spectrum. The second issue is that a simple system consisting only of a highly directional source transmitter and small optical detector receiver has a very narrow field of view. The goal of this project is to improve an optical, underwater communication system by increasing the effective field of view of the receiving optics. To this end, we make two changes to the simple system: (1) An optical dome was added near the receiver. An array of lenses is placed radially on the surface of the dome, reminiscent of the compound eye of an insect. The lenses make the source and detector planes conjugate, and each lens adds a new region of the source plane to the instrument's total field of view. (2) The receiver was expanded to include multiple photodiodes. With these two changes, the receiver has much more tolerance to misalignments (in position and angle) of the transmitter.Two versions of the optical dome (with 6 and 8 diameters) were designed using the CREO CAD software and modeled using the CODE V optical design software. A series of these transparent hemispherical domes, with both design diameters, were manufactured using a 5-axis mill. The prototype was then retrofitted with lenses and compared with the computer-generated model to demonstrate the effectiveness of this solution. This work shows the dome design improves the optical field of view of the underwater communication system considerably. Furthermore, with the experimental test results, a geometric optimization model was derived providing insights to the design performance limits.

Design↗

NASA NTRS2017

Hemispherical Optical Dome for Underwater Communication

For many years, acoustic systems have been used as the primary method for underwater communication; however, the data transfer rate of such systems is low because sound propagates slowly through the water. A higher throughput can be achieved using visible light to transmit data underwater. The first issue with using this approach is that there is generally a large loss of the light signal due to scattering and absorption in water even though there is an optimal wavelength for transmission in the blue or green wavelengths of the visible spectrum. The second issue is that a simple system consisting only of a highly directional source/transmitter and small optical detector/receiver has a very narrow field of view. The goal of this project is to improve an optical, underwater communication system by increasing the effective field of view of the receiving optics. To this end, we make two changes to the simple system: (1) An optical dome was added near the receiver. An array of lenses is placed radially on the surface of the dome, reminiscent of the compound eye of an insect. The lenses make the source and detector planes conjugate, and each lens adds a new region of the source plane to the instrument's total field of view. (2) The receiver was expanded to include multiple photodiodes. With these two changes, the receiver has much more tolerance to misalignments (in position and angle) of the transmitter.Two versions of the optical dome (with 6 and 8 diameters) were designed using the CREO CAD software and modeled using the CODE V optical design software. A series of these transparent hemispherical domes, with both design diameters, were manufactured using a 5-axis mill. The prototype was then retrofitted with lenses and compared with the computer-generated model to demonstrate the effectiveness of this solution. This work shows the dome design improves the optical field of view of the underwater communication system considerably. Furthermore, with the experimental test results, a geometric optimization model was derived providing insights to the design performance limits.

Underwater↗

NASA NTRS1991

Theoretical and experimental studies relevant to interpretation of auroral emissions

The accomplishments achieved over the past year are detailed with emphasis on the interpretation or auroral emissions and studies of potential spacecraft-induced contamination effects. Accordingly, the research was divided into two tasks. The first task is designed to add to the understanding of space vehicle induced external contamination. An experimental facility for simulation of the external environment for a spacecraft in low earth orbit was developed. The facility was used to make laboratory measurements of important phenomena required for improving the understanding of the space vehicle induced external environment and its effect on measurement of auroral emissions from space-based platforms. A workshop was sponsored to provide a forum for presentation of the latest research by nationally recognized experts on space vehicle contamination and to discuss the impact of this research on future missions involving space-based platforms. The second task is to add an ab initio auroral calculation to the extant ionospheric/thermospheric global modeling capabilities. Once the addition of the code was complete, the combined model was to be used to compare the relative intensities and behavior of various emission sources (dayglow, aurora, etc.). Such studies are essential to an understanding of the types of vacuum ultraviolet (VUV) auroral images which are expected to be available within two years with the successful deployment of the Ultraviolet Imager (UVI) on the ISTP POLAR spacecraft. In anticipation of this, the second task includes support for meetings of the science working group for the UVI to discuss operational and data analysis needs. Taken together, the proposed tasks outline a course of study designed to make significant contributions to the field of space-based auroral imaging.

Keffer, Charles E.↗

DOE OSTI2023

Virtual prototyping of liquid metal blanket performance in fusion pilot plant

Liquid metal blanket is a dominant design option for the next step fusion devices responsible for harvesting energy from fusion reaction, and simultaneously producing fuel for the same reaction through tritium breeding. Liquid metal blankets introduce additional complexity to the design due to fluid motion, fluid structure interaction, and magnetohydrodynamic (MHD) effects arising from the motion of the conducting fluid through the magnetic field. They are also directly affected by the plasma heat flux and neutronic fluence. PPPL is currently developing a virtual prototyping system for numerical analysis of the liquid metal blankets for future fusion devices. The system has a customized 3D computational fluid dynamics (CFD) code in its core, allowing MHD flow and conjugate heat transfer analysis in blankets fluids and solids. The code was successfully used before for dual coolant blanket analysis [A. Khodak et al., Fusion Eng. and Des. 137 (2018)]. Recently the same code was modified to allow verified simulation of MHD flows at high Hartmann numbers of several thousand typical for blanket applications. CFD code receives volumetric heat source distribution from the neutronic analysis based on MCNP code. In addition, direct tritium breeding simulation will be performed allowing optimization of the blanket performance. 2D axisymmetric version of neutronics code will be used for rapid optimization, with 3D version employed for detailed analysis. The surface heat distribution on the plasma facing wall will be defined by the software HEAT allowing 3D modeling of the heat flux based on the magnetic field distribution including gyro-orbit effects. Results of thermal analysis are imported into structural analysis code also included in the system. Finally, direct import of CAD geometry will be used for analyzing all components and as a result design option can be efficiently optimized.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

DOE OSTI2020

Impact of undoped substrates on high performance silicon solar cells

Today's highest-efficiency silicon solar cells typically operate near the threshold between low-level and high-level injection. It is not well understood if pushing further into a regime in which the cell operating point is solidly in high-level injection at all times of the day has further benefits for the solar cell performance. From a reliability perspective, cells fabricated on lower doped silicon have a larger breakdown voltage. This advantage can affect the design of modules allowing higher voltages and a relaxation of the number of bypass diodes needed. In this project, we present a comprehensive assessment, both experimental and using simulation, of how bulk resistivity, light intensity, and operation temperature impact the performance of the silicon solar cell. This work incorporates a comprehensive device physics analysis assisted by numerical simulation. The source code is now available under General Public License (GPL3), and to further leverage the findings of this project and outreach people outside of the scientific community, we are working with www.pveducation.org (which receives >1 million visitors a year) to create interactive content using our simulation results. The simulation results indicate that high bulk resistivity wafers (>>10 Ωcm) require bulk Shockley-Read-Hall (SRH) lifetimes in the millisecond range to outperform wafers with standard bulk resistivities (<10 Ωcm). Additionally, above bulk resistivities of 10 Ωcm (the exact value depends on the bulk characteristics of the wafer), the cell efficiency is weakly dependent on the bulk resistivity. As a result, ingot manufactures may have an opportunity to further reduce wafer cost by growing higher resistivity ingots that are more tolerant to resistivity variations. This project is particularly relevant today, as solar cell architectures with improved surface passivation and milliseconds lifetimes wafers are commercially available, leveraging potential benefits of using higher bulk resistivities. Outside of interdigitated back contact (IBC) cell, reported studies on high resistivity silicon (>100 Ωcm) are very limited. To the best of our knowledge, this project provides for the first-time experimental insight on solar cells fabricated on wafers with bulk resistivities up to several thousand Ωcm, delivering a comprehensive vision of their performance under real-world temperature and light intensity operation conditions. We manufactured and characterized solar cells with bulk resistivities in the range of 1 Ωcm to >15k Ωcm. Under standard testing conditions (STC), we measured solar cells efficiencies over 20% over the entire range of bulk resistivities, using our baseline cell processing. To evaluate the cell performance in real-world operation conditions, the solar cells were measured at different temperatures (25-80°C) and at different light intensities (0.1-1 suns). The measurements show that the bulk resistivity does not impact the solar cell response to temperature and light intensity. Similar thermal coefficients (TC) were measured for standard and high bulk resistivities, and they are comparable with the TC values reported in the literature for standard bulk resistivities <10 Ωcm. After light soaking, the solar cell didn’t show signs of light-induced degradation (LID). This result was expected since n-type float zone (FZ) wafers were used in this work, i.e. low traces of boron and low concentration of oxygen (oxygen is typically found in the seed end of Czochralski (CZ) ingots). We measured for high bulk resistivities (>10 Ωcm) extremely high breakdown voltages (>1000V). In conclusion, the insight provided by this project can positively impact the levelized cost of energy (LCOE) of the photovoltaic systems through its effect on cell and ingot manufacturing yield, silicon cell power output, and module reliability.

14 SOLAR ENERGY↗

NASA NTRS1991

Space nuclear power systems; Proceedings of the 8th Symposium, Albuquerque, NM, Jan. 6-10, 1991. Pts. 1-3

The present conference discusses NASA mission planning for space nuclear power, lunar mission design based on nuclear thermal rockets, inertial-electrostatic confinement fusion for space power, nuclear risk analysis of the Ulysses mission, the role of the interface in refractory metal alloy composites, an advanced thermionic reactor systems design code, and space high power nuclear-pumped lasers. Also discussed are exploration mission enhancements with power-beaming, power requirement estimates for a nuclear-powered manned Mars rover, SP-100 reactor design, safety, and testing, materials compatibility issues for fabric composite radiators, application of the enabler to nuclear electric propulsion, orbit-transfer with TOPAZ-type power sources, the thermoelectric properties of alloys, ruthenium silicide as a promising thermoelectric material, and innovative space-saving device for high-temperature piping systems. The second volume of this conference discusses engine concepts for nuclear electric propulsion, nuclear technologies for human exploration of the solar system, dynamic energy conversion, direct nuclear propulsion, thermionic conversion technology, reactor and power system control, thermal management, thermionic research, effects of radiation on electronics, heat-pipe technology, radioisotope power systems, and nuclear fuels for power reactors. The third volume discusses space power electronics, space nuclear fuels for propulsion reactors, power systems concepts, space power electronics systems, the use of artificial intelligence in space, flight qualifications and testing, microgravity two-phase flow, reactor manufacturing and processing, and space and environmental effects.

El-Genk, Mohamed S.↗

NASA NTRS2000

Flow Duct Data for Validation of Acoustic Liner Codes for Impedance Eduction

The objective of the study reported here was to acquire acoustic and flow data with hard and lined duct wall duct sections for validation of a liner prediction code being developed at NASA LaRC. Both the mean flowfield and acoustic flowfields were determined in a cross-plane of the rectangular duct. A flow duct facility with acoustic drivers connected to a rectangular (4.7 x 2.0 inch) source section and a linear acoustic liner mounted downstream of the source section was used in this study. The liner section was designed to allow liner materials to be placed on all 4 walls of the duct. The test liner was of the locally-reacting type and was made from a ceramic material. The material, consisting of a tubular structure, was provided by NASA LaRC. The liner was approximately 8.89 cm (3.5 inches) thick. For the current study, only the two "short" sides of the duct were lined with liner material. The other two sides were hard walls. Two especially built instrumentation sections were attached on either sides of the liner section to allow acoustic and flow measurements to be made upstream and downstream of the liner. The two instrumentation duct sections were built to allow measurement of acoustic and flow properties at planes perpendicular to flow upstream and downstream of the liner section. The instrumentation section was also designed to provide a streamwise gradient in acoustic (complex) pressure from which the acoustic particle velocity, needed for the model validation, can be computed. Flow measurements included pressure, temperature, and velocity profiles upstream of the liner section. The in-flow sound pressure levels and phases were obtained with a microphone probe equipped with a nose cone in two cross planes upstream of the liner and two cross plane downstream of the liner. In addition to the acoustic measurements at the cross planes. axial centerline acoustic data was acquired using an axially traversing microphone probe which was traversed from a location upstream of the liner to some distance downstream of the liner. All probes used here had to be calibrated with respect to a standard microphone equipped with a nose cone to allow for the effects of flow.

Ahuja, K. K.↗

NASA NTRS2003

Update on Integrated Optical Design Analyzer

Updated information on the Integrated Optical Design Analyzer (IODA) computer program has become available. IODA was described in Software for Multidisciplinary Concurrent Optical Design (MFS-31452), NASA Tech Briefs, Vol. 25, No. 10 (October 2001), page 8a. To recapitulate: IODA facilitates multidisciplinary concurrent engineering of highly precise optical instruments. The architecture of IODA was developed by reviewing design processes and software in an effort to automate design procedures. IODA significantly reduces design iteration cycle time and eliminates many potential sources of error. IODA integrates the modeling efforts of a team of experts in different disciplines (e.g., optics, structural analysis, and heat transfer) working at different locations and provides seamless fusion of data among thermal, structural, and optical models used to design an instrument. IODA is compatible with data files generated by the NASTRAN structural-analysis program and the Code V (Registered Trademark) optical-analysis program, and can be used to couple analyses performed by these two programs. IODA supports multiple-load-case analysis for quickly accomplishing trade studies. IODA can also model the transient response of an instrument under the influence of dynamic loads and disturbances.

Moore, James D., Jr.↗

DOE OSTI2021

Demonstration of the On-the-Fly Shielding Analysis Method: Spent Fuel and Waste Disposition

This report documents work performed supporting the US Department of Energy (DOE) Office of Nuclear Energy (NE) Spent Fuel and Waste Disposition (SFWD) Integrated Waste Management activities under work breakdown structure element 1.08.02.04.01, “Data and Tools Development, Validation, and Maintenance.” In particular, this report fulfills milestone M3SF-21OR020401016, “Implement on-the-fly dose analysis methodology in UNF-ST&DARDS” within work package SF-21OR02040101, “Commercial SNF Characterization - ORNL.” The Used Nuclear Fuel - Storage, Transportation & Disposal Analysis Resource and Data System (UNFST& DARDS) enables automated dose rate calculations for spent nuclear fuel (SNF) transportation packages and storage casks using a Monte Carlo radiation transport code. The explicit method uses a detailed model of the SNF system and its contents. Therefore, a dose rate calculation is required for each as-loaded transportation package or storage cask because the SNF assemblies within a canister typically have unique irradiation characteristics. An alternate method, referred to as the “on-the-fly” shielding analysis method, has been proposed that requires only a set of Monte Carlo dose rate calculations for each transportation packaging/storage cask design. The results of the Monte Carlo dose rate calculations are independent of the SNF assembly irradiation and decay characteristics. The dose rate values may then be combined with the radiation source strength of the SNF assemblies associated with a particular transportation packaging/storage cask design to determine actual dose rates. This report presents on-the-fly dose rate calculations for a representative SNF storage cask and verification of the on-the-fly dose rate calculation results by comparison with reference dose rate calculations using the explicit Monte Carlo dose rate calculation. The on-the-fly shielding analysis method was implemented in UNF-ST&DARDS. A Python program was developed to process the MAVRIC dose rate results obtained by source particle type, energy group, and fuel geometry region. A Python processor created binary files, which were saved as a special UNF-ST&DARDS library for on-the- fly shielding analyses. UNF-ST&DARDS uses the precalculated on-the-fly binary libraries generated by the Python data processor and directly executes the Python code for on-the-fly dose analysis. This Python code unzips the pre-generated binary files mentioned above, reads the data, and combines them with user-specified sources for dose and uncertainty calculations. The Python programs were verified using Excel calculations and by comparison with the values obtained with the MAVRIC post-processing utilities applied to the 3dmap files. This method can currently be used to determine dose rates for as-loaded HI-STORM FW storage casks. The UNF-ST&DARDS analysis wizard for on-the-fly shielding analysis is described in this report.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

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↗