Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “evaluation datasets”

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 1,063 records · Page 59
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

Stoichiometrically coupled carbon and nitrogen cycling in the MIcrobial-MIneral Carbon Stabilization model version 1.0 (MIMICS-CN v1.0)

Explicit consideration of microbial physiology in soil biogeochemical models that represent coupled carbon–nitrogen dynamics presents opportunities to deepen understanding of ecosystem responses to environmental change. The MIcrobial-MIneral Carbon Stabilization (MIMICS) model explicitly represents microbial physiology and physicochemical stabilization of soil carbon (C) on regional and global scales. Here we present a new version of MIMICS with coupled C and nitrogen (N) cycling through litter, microbial, and soil organic matter (SOM) pools. The model was parameterized and validated against C and N data from the Long-Term Inter-site Decomposition Experiment Team (LIDET; six litter types, 10 years of observations, and 13 sites across North America). The model simulates C and N losses from litterbags in the LIDET study with reasonable accuracy (C: R 2 =0.63; N: R 2 =0.29), which is comparable with simulations from the DAYCENT model that implicitly represents microbial activity (C: R 2 =0.67; N: R 2 =0.30). Subsequently, we evaluated equilibrium values of stocks (total soil C and N, microbial biomass C and N, inorganic N) and microbial process rates (soil heterotrophic respiration, N mineralization) simulated by MIMICS-CN across the 13 simulated LIDET sites against published observations from other continent-wide datasets. We found that MIMICS-CN produces equilibrium values in line with measured values, showing that the model generates plausible estimates of ecosystem soil biogeochemical dynamics across continental-scale gradients. MIMICS-CN provides a platform for coupling C and N projections in a microbially explicit model, but experiments still need to identify the physiological and stoichiometric characteristics of soil microbes, especially under environmental change scenarios.

58 GEOSCIENCES↗
DOE OSTI2023

Evaluating generative networks using Gaussian mixtures of image features

We develop a measure for evaluating the performance of generative networks given two sets of images. A popular performance measure currently used to do this is the Fréchet Inception Distance (FID). However, FID assumes that images featurized using the penultimate layer of Inception follow a Gaussian distribution. This assumption allows FID to be easily computed, since FID uses the 2-Wasserstein distance of two Gaussian distributions fitted to the featurized images. However, we show that Inception features of the ImageNet dataset are not Gaussian; in particular, each marginal is not Gaussian. To remedy this problem, we model the featurized images using Gaussian mixture models (GMMs) and compute the 2-Wasserstein distance restricted to GMMs. We define a performance measure, which we call WaM, on two sets of images by using inception (or another classifier) to featurize the images, estimate two GMMs, and use the restricted 2-Wasserstein distance to compare the GMMs. We experimentally show the advantages of WaM over FID, including how FID is more sensitive than WaM to image perturbations. By modelling the non-Gaussian features obtained from inception as GMMs and using a GMM metric, we can more accurately evaluate generative network performance.

machine learning, genrative adversarial networks↗
DOE OSTI2026

Transplatformer: translating toxicogenomic profiles between generations of platforms

Background Transcriptomic profiling technologies have advanced the analysis of biological and toxicological responses. However, substantial differences in probe design, dynamic range, gene coverage, and preprocessing pipelines across platforms introduce artifacts that limit cross-study integration and hinder the reuse of historical datasets. We aim to develop computational methods for accurate cross-platform translation to maximize the value of legacy resources. Results We present TransPlatformer a deep learning framework for translating gene expression profiles across heterogeneous toxicogenomics platforms. TransPlatformer employs a novel attention-based architecture to map high-dimensional fold-change vectors from legacy microarray technologies to current platforms. Models are trained and evaluated using DrugMatrix, spanning three technological generations. We investigate mixed-tissue, single-tissue, and cross-tissue training paradigms and benchmark performance against multilayer perceptron and matrix-completion baselines. In mixed-tissue training, TransPlatformer achieves a greater than 50% reduction in mean absolute error (0.043 vs. 0.09) and nearly doubles Pearson correlation ( ≈ 0.71 vs. 0.37) relative to baseline methods. Importantly, TransPlatformer preserves rare but biologically meaningful over- and under-expressed signals, with mean absolute error below 0.22. Single-tissue models yield further improvements for well-represented organs, such as a 10% reduction in liver mean absolute error, while underscoring the need for data augmentation strategies in low-sample tissues.ra Conclusions TransPlatformer provides an effective and scalable computational solution for cross-platform transcriptomic translation. By enabling biologically faithful harmonization of gene expression data, the proposed approach facilitates the reuse of legacy toxicogenomics datasets, enhances downstream biomarker discovery, and supports more reproducible predictive modeling in toxicology.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2021

A physics-informed and hierarchically regularized data-driven model for predicting fluid flow through porous media

This paper presents a new deep learning data-driven model for predicting structure dependent pore-fluid velocity fields in rock. The model is based on a Convolutional Auto-Encoder (CAE) artificial neural network capable of learning from image data generated by direct numerical simulations of fluid flow through pore-structures, such as by Lattice Boltzmann or molecular dynamics methods. The main novelty of the model in comparison to previous CAE-based data-driven approaches consists of three parts. The first is a methodology for decomposing the full-domain of the porous media into sub-regions, or “sub-domains”, in order to reduce the overall size of the CAE, batch process the sub-domains in parallel, and enable the CAE to learn local and generalizable nonlinear mappings of pore-fluid velocities. The second consists of embedding the finite difference solutions of the incompressible Navier-Stokes and continuity equations into convolutional layers prior to the CAE in order to provide the CAE with knowledge of fluid dynamics physics (PhyFlow). The third main novelty is that the training of the CAE is regularized with a hierarchical loss function that encourages the learning of fluid flow patterns (in a way similar to ranked modes in principal component analysis), ranking from most to least important. This is shown to increase the stability in learning, reduce over-fitting, and promote interpretability of the CAE neural network layers (HierCAE). The comprehensive new data-driven model, which we call the PhyFlow-HierCAE model, is shown to exhibit improved accuracy and generalizability of flow field predictions over conventional CAE models, attributable to the embedded physical knowledge and the hierarchical regularization, as well as realize orders of magnitude speed-ups in computation times as a surrogate for the direct numerical simulations. Examples of training and forward predictions on unseen pore-structures are provided and evaluated for data from Lattice Boltzmann and molecular dynamics simulations of pore-fluid flow. The model is shown to be a fast and accurate emulator (or “surrogate”) for predicting effective permeability of unseen pore-structures based on learning from relatively small direct numerical simulation datasets.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
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 OSTI2026

ChatMPI: LLM-Driven MPI Code Generation for HPC Workloads

The Message Passing Interface (MPI) standard plays a crucial role in enabling scientific applications for parallel computing and is an essential component in high-performance computing (HPC). However, implementing MPI code manually—especially applying a proper domain decomposition and communication pattern—is a challenging and error-prone task. We present ChatMPI, an AI assistant for MPI parallelization of sequential C codes. In our analysis, we focus on testing six essential HPC workloads, which are based on Basic Linear Algebra Subprograms levels 1, 2, and 3 as well as sparse, stencil, and iterative operations. We analyze the process of creating ChatMPI by using the ChatHPC library. This lightweight large language model (LLM)–based infrastructure enables HPC experts to efficiently create and supervise trustworthy AI capabilities for critical HPC software tasks. We study the data required for training (fine-tuning) ChatMPI to generate parallel codes that not only use MPI syntax correctly but also apply HPC techniques to reduce memory communication and maximize performance by using proper work decomposition. With a relatively small training dataset composed of a few dozen prompts and fewer than 15 minutes of fine-tuning on one node equipped with two NVIDIA H100 GPUs, ChatMPI elevates trustworthiness for MPI code generation of current LLMs (e.g., Code Llama, ChatGPT-4o and ChatGPT 5). Additionally, we evaluate the performance of the MPI codes generated by ChatMPI in comparison with the ones generated by ChatGPT-4o and ChatGPT-5. The codes generated by ChatMPI provide up to a 4 × boost in performance by using better problem decomposition, communication patterns, and HPC techniques (e.g., communication avoiding).

Valero Lara, Pedro [ORNL] (ORCID:0000000214794310)↗
DOE OSTI2023

Design and Development of a High Fidelity Cyber-Physical Testbed

In order to ensure that future critical infrastructure systems are resilient to various types of such advanced and persistent threats, it is important to develop and integrate tailored solutions that holistically address cyber-attack detection and mitigation in a timely manner such that adverse system impacts that impact a large population are avoided. Further, it is essential to create environments that allow control, protection and communication to exist within a realistic environment to analyze the effects of adverse conditions and system operating modes. This project aims to establish a high-fidelity testbed environment for modeling and simulating a single microgrid all the way up to a network of microgrids along with baseline controls, protection, and associated cyber communication. This is an important activity because accurately modeling and simulating the various power-electronics-based DERs and loads in a microgrid is critical to adequately capturing their behaviors over a wide range of off-normal conditions, as well as to evaluate the resilience of the system using the developed controls. The work presented in this report focuses on the process of building this high-fidelity testbed and the associated experimentation it enables. The model enables the creation of high-fidelity use cases and associated datasets that have been used extensively within the initiative to study resilience and support novel control development and prototyping. The work heavily leverages existing capability that is part of the high-fidelity cyber-physical system experimentation lab to create a power hardware-in-the-loop setup. The report also details the creation of an automated model building platform that can enable high-fidelity real-time models to be built without much effort allowing existing low-fidelity models to be analyzed in higher fidelity. Lastly, the report also discusses efforts center around scaling to large complex power system models to make the experimentation more effective.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Developing multi-gene CRISPRa/i programs to accelerate DBTL cycles in ABF hosts engineered for chemical production

This project developed and implemented a modular CRISPR activation and interference (CRISPRa/i) platform to accelerate strain optimization and pathway development for industrially relevant microbial hosts. By integrating multiplexed transcriptional perturbation tools with data-driven Design–Build–Test–Learn (DBTL) workflows, the team achieved reductions in cycle time and enhanced production of industrial aromatics, particularly 4-aminocinnamic acid (4-ACA), in Pseudomonas putida. Key accomplishments included: ● Development of a robust, tunable CRISPRa/i system in P. putida that enabled efficient multi-target gene regulation via guide RNA (gRNA) programs ● Completion of two full DBTL cycles, guided by machine learning (ML) models trained on transcriptomic and performance data, reducing engineering time by over 30% ● Optimization of multi-gene regulatory programs to balance expression of host and pathway modules, improve 4-ACA titers, and resolve metabolic bottlenecks ● Demonstration of system portability through a limited proof-of-concept extension in Acinetobacter baylyi, underscoring the generalizability of the approach ● Evaluation of strain performance on lignocellulosic biomass-derived substrates, demonstrating the feasibility of converting renewable carbon into aromatic building blocks These results illustrate the feasibility of applying ML-guided CRISPRa/i perturbation strategies to accelerate strain development in complex microbial systems. The resulting tools and datasets contribute to DOE objectives by improving platform predictability, reducing development costs, and enabling broader access to sustainable, economically viable bioproduction technologies.

09 BIOMASS FUELS↗
DOE OSTI2021

Omics-Lethal Human Viruses, Ebola Experiment EHUH002

The purpose of this experiment was to evaluate the human host response to Zaire Ebola wild-type virus and mutant virus infection. Samples were obtained from human hepatoma carcinoma cells (HUH-7) infected with Zaire Ebola ΔVP30-WT background for proteome, metabolome, and lipidome expression analysis. Secondary host-associated viral dataset downloads contain one or more statistically processed (normalization data transformation) quantitative dataset collections resulting in qualitative expression analyses of primary host-pathogen experimental study designs. Leveraging unique high-resolution Omics capabilities for proteomics, metabolomics, and lipidomics dataset downloads each have a direct relationship to a primary sample submission corresponding to a specific Ebola virus infection.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2021

Omics-Lethal Human Viruses, Influenza A Experiment IM101

The purpose of this experiment was to evaluate the host mouse response to wild-type Influenza A/Vietnam/1203/2004 (H5N1) virus, Influenza A/California/04/2009 (H1N1) virus, and Influenza A/Vietnam/1203/2004 (H5N1; "PB2-627E" and "NS1trunc124") infection. Sample data was obtained from whole mouse lung and processed for transcriptomics expression analysis. Secondary host-associated viral dataset downloads contain one or more statistically processed (normalization data transformation) quantitative dataset collections resulting in qualitative expression analyses of primary host-pathogen experimental study designs. Transcriptomics dataset downloads have a direct relationship to a primary sample submission corresponding to a specific Influenza A virus host infection.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2021

Omics-Lethal Human Viruses, West Nile Experiment WDC011

The purpose of this experiment was to evaluate the host response to wild-type West Nile virus (WNVMT) and mutant (WNVE218A) virus infection. Sample data was obtained from primary mouse myeloid dendritic cells for mRNA expression analysis. See Experiment WDC010 (GSE74628) for corresponding independent biological replicate study. Secondary host-associated viral dataset downloads contain one or more statistically processed (normalization data transformation) quantitative dataset collections resulting in qualitative expression analyses of primary host-pathogen experimental study designs. Transcriptomics dataset downloads have a direct relationship to a primary sample submission corresponding to a specific West Nile virus infection.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2023

Recurrent Convolutional Deep Neural Networks for Modeling Time-Resolved Wildfire Spread Behavior

The increasing incidence and severity of wildfires underscores the necessity of accurately predicting their behavior. While high-fidelity models derived from first principles offer physical accuracy, they are too computationally expensive for use in real-time fire response. Low-fidelity models sacrifice some physical accuracy and generalizability via the integration of empirical measurements, but enable real-time simulations for operational use in fire response. Machine learning techniques have demonstrated the ability to bridge these objectives by learning first-principles physics while achieving computational speedups. While deep learning approaches have demonstrated the ability to predict wildfire propagation over large time periods, time-resolved fire-spread predictions are needed for active fire management. Here, in this work, we evaluate the ability of deep learning approaches in accurately modeling the time-resolved dynamics of wildfires. We use an autoregressive process in which a convolutional recurrent deep learning model makes predictions that propagate a wildfire over 15 min increments. We apply the model to four simulated datasets of increasing complexity, containing both field fires with homogeneous fuel distribution as well as real-world topologies sampled from the California region of the United States. We show that even after 100 autoregressive predictions representing more than 24 h of simulated fire spread, the resulting models generate stable and realistic propagation dynamics, achieving a Jaccard score between 0.89 and 0.94 when predicting the resulting fire scar. The inference time of the deep learning models are examined and compared, and directions for future work are discussed.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2026

Quantifying Microstructure Variability in Laser Powder Bed Fusion 316 L Stainless Steel Microstructures with Spatial Statistics

Here, we have explored data-driven methods for material microstructure quantification that improve sensitivity to microstructural changes compared to traditional approaches. The methods integrate multiple microstructural properties, including grain morphology, crystallographic orientation, and material phase information. The simpler method employs maps of the Euclidean distance transformation metric to evaluate the morphology of grain boundary networks. The more intensive approach employs generalized spherical harmonic mapping for crystallographic orientations, per-pixel phase information, and a variational auto-encoder for dimensionality reduction and results in a multidimensional clustering of by microstructure similarity. Applied to an experimental dataset of additively manufactured steel, both methods detected slight variations in samples produced under nominally identical processing conditions. Both methods were able to distinguish between samples from multiple (nominally identical) builds, while the generalized spherical harmonics-based method could additionally cluster data samples rotated at two orientations on the build plate. The improved sensitivity of the methods, demonstrated through comparison with traditional microstructure characterization techniques, offers advantages for microstructure quantification and comparisons in advanced manufacturing applications.

SS316L↗
DOE OSTI2024

Predicting nutrition and environmental factors associated with female reproductive disorders using a knowledge graph and random forests

Female reproductive disorders (FRDs) are common health conditions that may present with significant symptoms. Diet and environment are potential areas for FRD interventions. We utilized a knowledge graph (KG) method to predict factors associated with common FRDs (for example, endometriosis, ovarian cyst, and uterine fibroids). We harmonized survey data from the Personalized Environment and Genes Study (PEGS) on internal and external environmental exposures and health conditions with biomedical ontology content. We merged the harmonized data and ontologies with supplemental nutrient and agricultural chemical data to create a KG. We analyzed the KG by embedding edges and applying a random forest for edge prediction to identify variables potentially associated with FRDs. We also conducted logistic regression analysis for comparison. Across 9765 PEGS respondents, the KG analysis resulted in 8535 significant or suggestive predicted links between FRDs and chemicals, phenotypes, and diseases. Amongst these links, 32 were exact matches when compared with the logistic regression results, including comorbidities, medications, foods, and occupational exposures. Mechanistic underpinnings of predicted links documented in the literature may support some of our findings. Our KG methods are useful for predicting possible associations in large, survey-based datasets with added information on directionality and magnitude of effect from logistic regression. These results should not be construed as causal but can support hypothesis generation. This investigation enabled the generation of hypotheses on a variety of potential links between FRDs and exposures. Future investigations should prospectively evaluate the variables hypothesized to impact FRDs.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2021

Exploring the impacts of conformer selection methods on ion mobility collision cross section predictions

The prediction of structure dependent molecular properties, such as collision cross sections as measured using ion mobility spectrometry, are crucially dependent on the selection of the correct population of molecular conformers. Here, we report an in-depth evaluation of multiple conformation selection techniques, including simple averaging, Boltzmann weighting, lowest energy selection, low energy threshold reductions, and similarity reduction. Generating 50,000 conformers each for 18 molecules, we used the In Silico Chemical Library Engine (ISiCLE) to calculate the collision cross sections for the entire dataset. First, we employed Monte Carlo simulations to understand the variability between conformer structures as generated using simulated annealing. Then we employed Monte Carlo simulations to the aforementioned conformer selection techniques applied on the simulated molecular property—the ion mobility collision cross section. Based on our analyses, we found Boltzmann weighting to be a good tradeoff between precision and theoretical accuracy. Combining multiple techniques revealed that energy thresholds and root-mean-squared deviation-based similarity reductions can save considerable computational expense while maintaining property prediction accuracy. Molecular dynamic conformer generation tools like AMBER can continue to generate new lowest energy conformers even after tens of thousands of generations, decreasing precision between runs. This reduced precision can be ameliorated and theoretical accuracy increased by running density functional theory geometry optimization on carefully selected conformers.

Nielson, Felicity F.↗
DOE OSTI2022

Identification of beneficial and detrimental bacteria impacting sorghum responses to drought using multi-scale and multi-system microbiome comparisons

Abstract Drought is a major abiotic stress limiting agricultural productivity. Previous field-level experiments have demonstrated that drought decreases microbiome diversity in the root and rhizosphere. How these changes ultimately affect plant health remains elusive. Toward this end, we combined reductionist, transitional and ecological approaches, applied to the staple cereal crop sorghum to identify key root-associated microbes that robustly affect drought-stressed plant phenotypes. Fifty-three Arabidopsis-associated bacteria were applied to sorghum seeds and their effect on root growth was monitored. Two Arthrobacter strains caused root growth inhibition (RGI) in Arabidopsis and sorghum. In the context of synthetic communities, Variovorax strains were able to protect plants from Arthrobacter-caused RGI. As a transitional system, high-throughput phenotyping was used to test the synthetic communities. During drought stress, plants colonized by Arthrobacter had reduced growth and leaf water content. Plants colonized by both Arthrobacter and Variovorax performed as well or better than control plants. In parallel, we performed a field trial wherein sorghum was evaluated across drought conditions. By incorporating data on soil properties into the microbiome analysis, we accounted for experimental noise with a novel method and were able to observe the negative correlation between the abundance of Arthrobacter and plant growth. Having validated this approach, we cross-referenced datasets from the high-throughput phenotyping and field experiments and report a list of bacteria with high confidence that positively associated with plant growth under drought stress. In conclusion, a three-tiered experimental system successfully spanned the lab-to-field gap and identified beneficial and deleterious bacterial strains for sorghum under drought.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

Extreme flood return levels in a U.S. mid-Atlantic estuary using 40-year fluvial-coastal model simulations

Using an integrated watershed-coastal modeling framework, we conducted long-term historical simulations (1980-2019) of fluvial and coastal flooding in the Delaware Bay and River, a vulnerable estuarine system in the U.S., at high spatial resolutions. By applying process-based models, we generated physically consistent and spatially detailed estimates of estuarine, riverine, and surge-driven extreme water level compared to previous studies that used field data only. We then evaluated changes in the magnitude of flood events using the 40-year simulations and detrended Floodwater Depth values with stationary extreme value analysis. Our detailed assessment of spatial-varying extreme values revealed how different flood-generation mechanisms can dominate various zones in the estuary. The datasets produced through this work will be valuable for long-term flood hazard mitigation planning in coastal communities in the Delaware Bay and River region. Additionally, this work will serve as a benchmark for other coastal flood hazard modeling communities worldwide, aiding them in systematically modeling long-term and continuous extreme flood events.

Coastal and Riverine Flooding↗
DOE OSTI2025

Data-driven emulation of modal aerosol microphysics via neural operator-based modeling

The complexity and the small characteristic scales of aerosol microphysical processes pose a big challenge for accurate and efficient Earth system simulations at regional and global scales. In this work, we construct and evaluate a surrogate model: the aerosol deep operator network (ADON), a physics-inspired dual-net architecture for emulating the aerosol microphysics parameterization suite in the version 2 of the Energy Earth System Model (E3SMv2). The current version of the surrogate model is trained on a dataset comprising 9.8 million samples obtained from a global E3SMv2 simulation with the horizontal resolution of about one degree under cloud-free conditions. Incorporating domain spatial and temporal coordinates, as well as principle components extracted from training data, the dual-net surrogate model effectively captures the intricate representations of aerosol and the relationship with atmospheric state variables, achieving an R-squared score over $$95.7\%$$ for all the lognormal aerosol modes in the extrapolated regime. The validated model provides feature importance of input variables and their impact on the predictive capacity of the surrogate model in relation to the E3SM. The computational cost of online inference time deployed on CPUs and GPUs with lower precisions highlights ADON’s efficiency and potential in robust predictive modeling for large-scale Earth system computations.

Bai, Zhe↗