Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Model error”

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 1,657 records · Page 92
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

Self-diagnosis of model suitability for continuous measurements of stream-dissolved organic carbon derived from in situ UV–visible spectroscopy

Application of high-frequency monitoring of dissolved organic carbon (DOC) is difficult in instances where training datasets are challenging to develop (e.g., remote locations) and the relationship between optical features and DOC concentration changes due to environmental or landscape shifts (e.g., climate or land-use change). We developed and compared three partial least squares (PLS) models using in situ water level measurements, conductivity, and UV–Vis spectral attenuation to predict DOC. Two site-specific models were developed using data from a hillslope-dominated forest or a low-relief wetland-pond-dominated stream catchment. The third model, using data from both sites, exhibited the best performance (DOC range = 4–15.5 mg C L −1 , mean = 8.38 mg C L −1 , training RMSE = 0.34 mg C L −1 , internal validation RMSE = 0.50 mg C L −1 , external validation RMSE = 2.43 mg C L −1 ). We further demonstrate using PLS model statistics to monitor performance and elucidate when and how models should be updated. These statistics, Hotelling's T 2 and squared prediction errors, are useful consistency checks for the predictions made and detect underlying inconsistencies that, if undetected, can reduce the robustness of DOC prediction. For example, via the T 2 statistic, we identified the summer–autumn transition as a period when DOC composition differed from what was represented in the training dataset. We also determined that elevated SUVA 254 values contributed to the overall bias observed in predictions made during the subsequent year as part of the external validation. This enabled the application of a bias correction that reduced the RMSE from 2.43 to 0.89 mg C L −1 . The method presented here could be applied to future monitoring programs enabling model updates to monitor DOC fluxes accurately from optical datasets (e.g., attenuance or fluorescence) in the face of developing datasets in remote locations or environmental change. In conclusion, implementation of this approach may also identify possible regime shifts or landscape and hydrologic change associated with climate and other environmental changes relevant to terrestrial to aquatic fluxes.

UV-VIS spectroscopy↗
DOE OSTI2023

CrossLink: Geometry API [Slides]

The mesh generation process is very challenging and time consuming when working with complex CAD models. The process of creating and sorting geometric entities into groups appropriate for meshing is labor intensive and prone to error. In addition, the common data exchange formats such as STEP and IGES do not propagate information such as entity names that may be defined in the original model. Finally, entity counts change frequently with parameter variation as a result of tolerance-based geometry operations. Thus, sorting by index does not provide a robust and repeatable means for grouping. xGeom is a geometry library that enables the creation of NURBS curves and surfaces via a python scripting interface. xGeom is ideal for studying relatively simple models and is fully integrated with CrossLink’s mesh generation capabilities. For more complex models, xCAD is a python-based Creo Parametric CAD model driver that enables the model to be generated, queried, parametrically modified, regenerated, and exported without data loss and in a fully repeatable manner.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Dynamic global vegetation models underestimate net CO 2 flux mean and inter-annual variability in dryland ecosystems

Despite their sparse vegetation, dryland regions exert a huge influence over global biogeochemical cycles because they cover more than 40% of the world surface (Schimel 2010 Science 327 418–9). It is thought that drylands dominate the inter-annual variability (IAV) and long-term trend in the global carbon (C) cycle (Poulter et al 2014 Nature 509 600–3, Ahlstrom et al 2015 Science 348 895–9, Zhang et al 2018 Glob. Change Biol. 24 3954–68). Projections of the global land C sink therefore rely on accurate representation of dryland C cycle processes; however, the dynamic global vegetation models (DGVMs) used in future projections have rarely been evaluated against dryland C flux data. Here, we carried out an evaluation of 14 DGVMs (TRENDY v7) against net ecosystem exchange (NEE) data from 12 dryland flux sites in the southwestern US encompassing a range of ecosystem types (forests, shrub- and grasslands). We find that all the models underestimate both mean annual C uptake/release as well as the magnitude of NEE IAV, suggesting that improvements in representing dryland regions may improve global C cycle projections. Across all models, the sensitivity and timing of ecosystem C uptake to plant available moisture was at fault. Spring biases in gross primary production (GPP) dominate the underestimate of mean annual NEE, whereas models' lack of GPP response to water availability in both spring and summer monsoon are responsible for inability to capture NEE IAV. Errors in GPP moisture sensitivity at high elevation forested sites were more prominent during the spring, while errors at the low elevation shrub and grass-dominated sites were more important during the monsoon. We propose a range of hypotheses for why model GPP does not respond sufficiently to changing water availability that can serve as a guide for future dryland DGVM developments. Our analysis suggests that improvements in modeling C cycle processes across more than a quarter of the Earth's land surface could be achieved by addressing the moisture sensitivity of dryland C uptake.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2021

Adaptive resource allocation for surrogate modeling of systems comprised of multiple disciplines with varying fidelity

We present an adaptive algorithm for constructing surrogate models for integrated systems composed of a set of coupled components. With this goal we introduce ‘coupling’ variables with a priori unknown distributions that allow approximations of each component to be built independently. Once built, the surrogates of the components are combined and used to predict system-level quantities of interest (QoI) at a fraction of the cost of interrogating the full system model. We use a greedy experimental design procedure, based upon a modification of Multi-Index Stochastic Collocation (MISC), to minimize the error of the combined surrogate. This is achieved by refining each component surrogate in accordance with its relative contribution to error in the approximation of the system-level QoI. Our adaptation of MISC is a multi-fidelity procedure that can leverage ensembles of models of varying cost and accuracy, for one or more components, to produce estimates of system-level QoI. Several numerical examples demonstrate the efficacy of the proposed approach on systems involving feed-forward and feedback coupling. For a fixed computational budget, the proposed algorithm is able to produce approximations that are orders of magnitude more accurate than approximations that treat the integrated system as a black-box.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2020

Can machine learning improve the model representation of turbulent kinetic energy dissipation rate in the boundary layer for complex terrain?

Current turbulence parameterizations in numerical weather prediction models at the mesoscale assume a local equilibrium between production and dissipation of turbulence. As this assumption does not hold at fine horizontal resolutions, improved ways to represent turbulent kinetic energy (TKE) dissipation rate (ϵ) are needed. Here, we use a 6-week data set of turbulence measurements from 184 sonic anemometers in complex terrain at the Perdigão field campaign to suggest improved representations of dissipation rate. First, we demonstrate that the widely used Mellor, Yamada, Nakanishi, and Niino (MYNN) parameterization of TKE dissipation rate leads to a large inaccuracy and bias in the representation of ϵ. Next, we assess the potential of machine-learning techniques to predict TKE dissipation rate from a set of atmospheric and terrain-related features. We train and test several machine-learning algorithms using the data at Perdigão, and we find that the models eliminate the bias MYNN currently shows in representing ϵ, while also reducing the average error by up to almost 40 %. Of all the variables included in the algorithms, TKE is the variable responsible for most of the variability of ϵ, and a strong positive correlation exists between the two. These results suggest further consideration of machine-learning techniques to enhance parameterizations of turbulence in numerical weather prediction models.

58 GEOSCIENCES↗
DOE OSTI2022

Accelerating error correction in tomographic reconstruction

Abstract Spurred by recent advances in detector technology and X-ray optics, upgrades to scanning-probe-based tomographic imaging have led to an exponential growth in the amount and complexity of experimental data and have created a clear opportunity for tomographic imaging to approach single-atom sensitivity. The improved spatial resolution, however, is highly susceptible to systematic and random experimental errors, such as center of rotation drifts, which may lead to imaging artifacts and prevent reliable data extraction. Here, we present a model-based approach that simultaneously optimizes the reconstructed specimen and sinogram alignment as a single optimization problem for tomographic reconstruction with center of rotation error correction. Our algorithm utilizes an adaptive regularizer that is dynamically adjusted at each alternating iteration step. Furthermore, we describe its implementation in a software package targeting high-throughput workflows for execution on distributed-memory clusters. We demonstrate the performance of our solver on large-scale synthetic problems and show that it is robust to a wide range of noise and experimental drifts with near-ideal throughput.

Ali, Sajid (ORCID:0000000321864636)↗
DOE OSTI2025

Enabling Robust Exoplanet Atmospheric Retrievals with Gaussian Processes

Atmospheric retrievals are essential tools for interpreting exoplanet transmission and eclipse spectra, enabling quantitative constraints on the chemical composition, aerosol properties, and thermal structure of planetary atmospheres. The James Webb Space Telescope (JWST) offers unprecedented spectral precision, resolution, and wavelength coverage, unlocking transformative insights into the formation, evolution, climate, and potential habitability of planetary systems. However, this opportunity is accompanied by challenges: modeling assumptions and unaccounted-for noise or signal sources can bias retrieval outcomes and their interpretation. To address these limitations, we introduce a Gaussian process (GP)-aided atmospheric retrieval framework that flexibly accounts for unmodeled features and correlated noise in exoplanet spectra. We validate this method on synthetic JWST observations, and show that GP-aided retrievals reduce bias in inferred abundances and better capture model–data mismatches than traditional approaches. We also introduce the concept of mean squared error to quantify the trade-off between bias and variance, arguing that this metric more accurately reflects retrieval performance than bias alone. We then reanalyze the NIRISS/SOSS JWST transmission spectrum of WASP-96 b, finding that GP-aided retrievals yield broader constraints on CO 2 and H 2 O, possibly alleviating tension between previous retrieval results and equilibrium predictions. Our GP framework provides precise and accurate constraints while highlighting regions where models fail to explain the data. As JWST matures and future facilities come online, a deeper understanding of the limitations of both data and models will be essential, and GP-enabled retrievals like the one presented here offer a principled path forward.

Rotman, Yoav [Arizona State Univ., Tempe, AZ (Unit↗
DOE OSTI2026

HydraGNN_Predictive_GFM_2026 - Ensemble of predictive graph foundation models for atomistic materials modeling

This release contains data and parameters of HydraGNN-based graph foundation models trained as a result of the work published in the pre-print "Exascale Multi-Task Graph Foundation Models for Imbalanced, Multi-Fidelity Atomistic Data" by M. Lupo Pasini et al. (https://arxiv.org/abs/2604.15380). We jointly train on 16 open first-principles datasets (544+ million structures covering 85+ elements) using a multi-task architecture with per-dataset heads and a scalable ADIOS2/DDStore data pipeline. On Frontier, we execute six large-scale DeepHyper hyperparameter optimization campaigns in FP64 and promote the top-performing message-passing models to sustained 2,048-node training, yielding a PaiNN-based lead model. The version of HydraGNN used to generate the outputs provided in this release is HydraGNN v5.0 (https://github.com/ORNL/HydraGNN/releases/tag/v5.0) The list of datasets used for the training of the graph foundation model is the following: 1) Alexandria [1] 2) ANI1x [2] 3) MPTrj [3] 4) Open Catalyst 2020 (OC20) [4] 5) Open Catalyst 2022 (OC22) [5] 6) Open Catalyst 2025 (OC25) [6] 7) Open Direct ir Capture 2023 (ODAC23) [7] 8) Open Materials 2024 (OMat24) [8] 9) Open Molecules 2025 (OMol25) [9] 10) OMol25-neutral (subset of OMol25 that contains only molecules with zero total charge) 11) OMol25-non-neutral (subset of OMol25 that contains only molecules with non-zero total charge) 12) Open Polymers 2026 (OPoly2026) [10] 13) Nabla2DFT [11] 14) QCML [12] 15) QM7X [reference 13] 16) transition1x [14] Dataset references: [1] J. Schmidt et al., “A dataset of 175k stable and metastable materials calculated with the PBEsol and SCAN functionals,” Scientific Data, vol. 9, p. 64, 2022. [2] J. S. Smith et al., “The ANI-1ccx and ANI-1x data sets, coupled-cluster and density functional theory properties for molecules,” Scientific Data, vol. 7, p. 134, 2020. [Online]. Available: https: //www.nature.com/articles/s41597-020-0473-z [3] A. Jain et al., “Commentary: The Materials Project: A materials genome approach to accelerating materials innovation,” APL Materials, vol. 1, no. 1, p. 011002, 07 2013. [Online]. Available: https://doi.org/10.1063/1.4812323 [4] L. Chanussot et al., “Open catalyst 2020 (oc20) dataset and community challenges,” ACS Catalysis, vol. 11, no. 10, pp. 6059–6072, 2021. [Online]. Available: https://doi.org/10.1021/acscatal.0c04525 [5] K. Tran et al., “Open catalyst 2022 (oc22) dataset and challenges for oxidation electrocatalysts,” ACS Catalysis, vol. 13, no. 5, pp. 3066–3084, 2023. [Online]. Available: https://doi.org/10.1021/acscatal.2c05426 [6] S. J. Sahoo et al., “The open catalyst 2025 (oc25) dataset and models for solid-liquid interfaces,” arXiv preprint arXiv:2509.17862, 2025. [Online]. Available: https://arxiv.org/abs/2509.17862 [7] A. Sriram et al., “The open DAC 2023 dataset and challenges for sorbent discovery in direct air capture,” ACS Central Science, vol. 10, no. 5, pp. 923–941, 2024. [8] L. Barroso-Luque et al., “Open materials 2024 (omat24) inorganic materials dataset and models,” 2024. [Online]. Available: https://arxiv.org/abs/2410.12771 [9] D. S. Levine et al., “The open molecules 2025 (OMol25) dataset, evaluations, and models,” 2025. [Online]. Available: https://arxiv.org/abs/2505.08762 [10] D. S. Levine et al., The open polymers 2026 (OPoly26) dataset and evaluations,” arXiv preprint arXiv:2512.23117, 2025. [Online]. Available: https://arxiv.org/abs/2512.23117 [11] K. Khrabrov et al., “Nabla2dft: A universal quantum chemistry dataset of drug-like molecules and a benchmark for neural network potentials,” in NeurIPS 2024 Datasets and Benchmarks Track, 2024. [Online]. Available: https://openreview.net/forum?id=ElUrNM9U8c [12] S. Ganscha et al., “The QCML dataset, quantum chemistry reference data from 33.5M DFT and 14.7B semi-empirical calculations,” Scientific Data, vol. 12, p. 406, 2025. [13] J. Hoja et al., “QM7-X, a comprehensive dataset of quantum-mechanical properties spanning the chemical space of small organic molecules,” Scientific Data, vol. 8, p. 43, 2021. [Online]. Available: https://www.nature.com/articles/s41597-021-00812-2 [14] M. Schreiner et al., “Transition1x - a dataset for building generalizable reactive machine learning potentials,” Scientific Data, vol. 9, p. 779, 2022. The folder "datasets_ADIOS2_format" contains the set of pre-processed datasets in Adaptable I/O System (ADIOS) format (https://www.exascaleproject.org/research-project/adios/) that have been used for the development and training of GFMs in this work. The "datasets_ADIOS2_format" directory contains 2 sub-directories, one for the version "v1" of the datasets and one for the version "v2" of the datasets. The version "v1" of the datasets provides values of the total energy as they are extracted from the original data as it was released by the respective institutions. The version "v2" of the datasets provides values of the energy that have been realigned. The realignment was performed by training a linear regression model that predicts the total energy as a function of the chemical composition of the atomistic structure, and then subtract such prediction from the original value of the total energy. Both folders "v1" and "v2" contain 16 sub-directories, each corresponding to an ADIOS2-formatted dataset The folder "DeepHyper-results" contains the configurational files and model's parameters for all the 186 HPO trials that were successfully completed by the scalable hyperparameter optimization (HPO) runs on Frontier. The content of the folder "DeepHyper-results" I structured as follows: 1) task-list.txt: list of mpnn name, jobid, and deephyper task id 2) gfm_${MPNN}_${JOBID}_0.${TASKID}: run directory with checkpoint files 3) gfm_${MPNN}: deephyper summary directory (*.csv) for each specific MPNN type 4) deephyper-experiment-${JOBID}: output and error logs for each job The file "deephyper-sorted.csv" contains the details of each HydraGNN model built and tested by HPO, obtained by merging the (*.csv) filed from each HPO run executed. Out of all the HPO trials, we selected 10 to continue the training of the respective HydraGNN models. Due to limited computational budget available in the LRN070 allocation we could not complete the training till convergence for all these 10 selected models. The folder "models" contains multiple sub-folders, one per each HydraGNN model trained. Each model sub-folder contains the parameters of each HydraGNN model, with multiple checkpoint-restarts. The list of sub-folders are as follows: 1) multidataset_hpo-BEST1-fp64 2) multidataset_hpo-BEST2-fp64 3) multidataset_hpo-BEST3-fp64 4) multidataset_hpo-BEST4-fp64 5) multidataset_hpo-BEST5-fp64 6) multidataset_hpo-BEST6-fp64 7) multidataset_hpo-BEST7-fp64 8) multidataset_hpo-BEST8-fp64 9) multidataset_hpo-BEST9-fp64 10) multidataset_hpo-BEST10-fp64 Within each one of these folders, additional auxiliary log files are provided with descriptions about how the training proceeded. The lead PaiNN-model is contained inside "multidataset_hpo-BEST6-fp64". The file "mlp_branch_weights" contains the parameters of the multi-layer perceptron (MLP) used to reconcile the predictions of the 16 output decoding heads of the HydragNN architectures. The MLP takes in input the chemical composition of the atomistic structure and predicts averaging weights to linearly mix the predictions of each output decoding head toward consolidating them into a single one. The folder "1.1billion-structure-inference" contains 1.1 billion atomistic structures randomly generated. Each structures is associated with energy and forces predicted with the lead-PaiNN model combined with the MLP model for reconciliation of the multi-branch predictions generated by the 16 output decoding heads. The folder "1.1billion-structure-inference" contains 9,300 (*.tar.gz) subdirectories, one per Frontier compute node used to execute the inference at exascale. Once uncompressed, each (*.tar.gz) subdirectory contains an ADIOS2 (*.bp) file container, where each atomistic structure is stored as a PyTorch-Geometric Data object. The file "export_dataset_environment_variables.sh" contains the environment variables that need to be set before running the HydraGNN code to reproduce the results provided in this dataset release. The code that can be used to load the ADIOS2 files, load HydraGNN models, and run inference is available at: https://github.com/ORNL/HydraGNN/releases/tag/v5.0

36 MATERIALS SCIENCE↗
DOE OSTI2026

A Faster-Than-Real-Time Framework for Reliability-Oriented Simulation of PV Inverters

Physics-of-Failure (PoF) based reliability assessment for photovoltaic (PV) inverters requires long-duration electrical and electrothermal stress histories, yet generating such stress histories with high-fidelity switching models over year long mission profiles is computationally prohibitive. Conventional methods either sacrifice modeling fidelity for speed or require runtimes that are impractical for design iteration and uncertainty studies. To address this bottleneck, this paper presents a High-Performance Computing (HPC) based simulation frame work for faster-than-real-time reliability-oriented simulation. The proposed framework integrates the Average-to-Switching (A2S) method with parallel computing techniques to accelerate switching-level waveform reconstruction. We further introduce optimization strategies, including cluster merging and sensitivity based mission profile screening, to reduce the computational burden. Evaluated using real-world mission profile inputs and a MATLAB/Simulink switching-model reference, the framework reduces the simulation time for a one-year mission from an intractable multi-year duration to approximately 7.3 minutes while maintaining low waveform error. This acceleration provides a practical reliability-oriented simulation engine that can be coupled with component-specific aging models for subsequent PV inverter PoF assessment.

High-performance Computing↗
DOE OSTI2024

Assessing the reliability of medical resource demand models in the context of COVID-19

Abstract Background Numerous medical resource demand models have been created as tools for governments or hospitals, aiming to predict the need for crucial resources like ventilators, hospital beds, personal protective equipment (PPE), and diagnostic kits during crises such as the COVID-19 pandemic. However, the reliability of these demand models remains uncertain. Methods Demand models typically consist of two main components: hospital use epidemiological models that predict hospitalizations or daily admissions, and a demand calculator that translates the outputs of the epidemiological model into predictions for resource usage. We conducted separate analyses to evaluate each of these components. In the first analysis, we validated various hospital use epidemiological models using a recent validation framework designed for epidemiological models. This allowed us to quantify the accuracy of the models in predicting critical aspects such as the date and magnitude of local COVID-19 peaks, among other factors. In the second analysis, we evaluated a range of demand calculators for ventilators, medical gowns, and COVID-19 test kits. To achieve this, we decoupled these demand calculators from the underlying epidemiological models and provided ground truth data for their inputs. This approach enabled a direct comparison of the demand calculators, comparing them against each other and actual usage data when available. The code is available athttps://doi.org/10.5281/zenodo.13712387. Results Performance varied greatly across the epidemiological models, with greater variability in COVID-19 hospital use predictions than for COVID-19 deaths as analyzed previously. Some models did not have any peaks. Among those that did, the models under-estimated date of peak approximately as often as they over-estimated, but were more likely to under-estimate magnitude of peak, with typical relative errors around 50%. Regarding demand calculator predictions, there was significant variability, including five-fold differences in predictions for gown models. Validation against actual or surrogate usage data illustrated the potential value of demand models while demonstrating their limitations. Conclusions The emerging field of demand modeling holds promise in averting medical resource shortages during future public health emergencies. However, achieving this potential necessitates focused efforts on standardization, transparency, and rigorous model validation before placing reliance on demand models in critical public health decision-making.

Medical Informatics↗
DOE OSTI2024

Addressing Issues with Working Memory in Video Object Segmentation

Contemporary state-of-the-art video object segmentation (VOS) models compare incoming unannotated images to a history of image-mask relations via affinity or cross-attention to predict object masks. We refer to the internal memory state of the initial image-mask pair and past image-masks as a working memory buffer. While the current state of the art models perform very well on clean video data, their reliance on a working memory of previous frames leaves room for error. Affinity-based algorithms include the inductive bias that there is temporal continuity between consecutive frames. To account for inconsistent camera views of the desired object, working memory models need an algorithmic modification that regulates the memory updates and avoid writing irrelevant frames into working memory. A simple algorithmic change is proposed that can be applied to any existing working memory-based VOS model to improve performance on inconsistent views, such as sudden camera cuts, frame interjections, and extreme context changes. The resulting model performances show significant improvement on video data with these frame interjections over the same model without the algorithmic addition. Our contribution is a simple decision function that determines whether working memory should be updated based on the detection of sudden, extreme changes and the assumption that the object is no longer in frame. By implementing algorithmic changes, such as this, we can increase the real-world applicability of current VOS models.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Toward Hybrid Physics-Machine Learning to improve Land Surface Model predictions

A critical challenge for Land Surface Models (LSMs) is to simulate processes at the surface and the subsurface and their feedbacks to the atmosphere. Even using the same climate forcings, different LSMs predict different surface fluxes and soil moisture conditions due to differences in the formulations of individual processes, parameterizations, and representation of spatial heterogeneity. Ultimately, these differences contribute to the LSM prediction errors and uncertainty. This research seeks to address this challenge by coupling physics-based modeling with state-of-the-art machine learning (ML) techniques to describe complex physical and biogeochemical processes and narrow the gap between model predictions and observations.

58 GEOSCIENCES↗
DOE OSTI2021

Detecting Arsenic Contamination Using Satellite Imagery and Machine Learning

Arsenic, a potent carcinogen and neurotoxin, affects over 200 million people globally. Current detection methods are laborious, expensive, and unscalable, being difficult to implement in developing regions and during crises such as COVID-19. This study attempts to determine if a relationship exists between soil’s hyperspectral data and arsenic concentration using NASA’s Hyperion satellite. It is the first arsenic study to use satellite-based hyperspectral data and apply a classification approach. Four regression machine learning models are tested to determine this correlation in soil with bare land cover. Raw data are converted to reflectance, problematic atmospheric influences are removed, characteristic wavelengths are selected, and four noise reduction algorithms are tested. The combination of data augmentation, Genetic Algorithm, Second Derivative Transformation, and Random Forest regression (R 2 =0.840 and normalized root mean squared error (re-scaled to [0,1]) = 0.122) shows strong correlation, performing better than past models despite using noisier satellite data (versus lab-processed samples). Three binary classification machine learning models are then applied to identify high-risk shrub-covered regions in ten U.S. states, achieving strong accuracy (=0.693) and F1-score (=0.728). Overall, these results suggest that such a methodology is practical and can provide a sustainable alternative to arsenic contamination detection.

63 RADIATION, THERMAL, AND OTHER ENVIRON. POLLUTAN↗
DOE OSTI2021

CrossCheck: Rapid, Reproducible, and Interpretable Model Evaluation

Evaluation beyond aggregate performance metrics, e.g. F1-score, is crucial to both establish an appropriate level of trust in machine learning models and identify future model improvements. In this paper we demonstrate CrossCheck, an interactive visualization tool for rapid crossmodel comparison and reproducible error analysis. We describe the tool and discuss design and implementation details. We then present three use cases (named entity recognition, reading comprehension, and clickbait detection) that show the benefits of using the tool for model evaluation. CrossCheck allows data scientists to make informed decisions to choose between multiple models, identify when the models are correct and for which examples, investigate whether the models are making the same mistakes as humans, evaluate models’ generalizability and highlight models’ limitations, strengths and weaknesses. Furthermore, CrossCheck is implemented as a Jupyter widget, which allows rapid and convenient integration into data scientists’ model development workflows.

Arendt, Dustin L.↗
DOE OSTI2025

Hierarchical Bayesian Inverse Problems: A High-Dimensional Statistics Viewpoint

This paper analyzes hierarchical Bayesian inverse problems using techniques from highdimensional statistics. Furthermore, our analysis leverages a property of hierarchical Bayesian regularizers that we call approximate decomposability to obtain non-asymptotic bounds on the reconstruction error attained by maximum a posteriori estimators. The new theory explains how hierarchical Bayesian models that exploit sparsity, group sparsity, and sparse representations of the unknown parameter can achieve accurate reconstructions in high-dimensional settings.

MAP estimation↗
DOE OSTI2025

A 1 km soil moisture dataset over eastern CONUS generated by assimilating SMAP data into the Noah-MP land surface model

An improved fine-scale soil moisture (SM) dataset at 1 km grid spacing, covering much of the eastern continental US, was generated by assimilating 9 km Soil Moisture Active Passive (SMAP) SM data into the v4.0.1 Noah-MP land surface model. With 12 ensemble members, the assimilation was carried out using the ensemble Kalman filter algorithm within NASA's Land Information System. The SM analysis for 2016 was fully validated against in situ observations from four different networks and compared with four other existing datasets. Results indicate that this SM analysis surpasses other datasets in top-layer SM distribution, including a machine-learning-based product, despite all SM estimates being less heterogeneous than observed. The analysis of anomalous errors suggests that large similarity in intrinsic errors is likely due to overlapping data sources among the selected SM datasets. More detailed evaluations were performed over two geographic areas. The observations collected by the Atmospheric Radiation Measurement facility in Oklahoma suggest that soil temperature and surface heat fluxes are concurrently simulated with good accuracy. Investigation into the 2016 southeastern US drought response further indicates drier conditions and higher evapotranspiration estimates compared to GLEAMv4.1. Notably, large errors are associated with grids having clay soil textures, underscoring the need for refined model treatments for specific soil types to further improve SM estimates. The dataset is publicly available on Zenodo at https://doi.org/10.5281/zenodo.14370563 (Tai et al., 2024).

Tai, Sheng-Lun [Pacific Northwest National Laborat↗
DOE OSTI2026

Deep learning–based digital twins for heat pumps

Heat pumps are effective cooling and heating appliances to save energy in buildings. However, traditional heat pump models are challenging to integrate with building demands in a co-simulation environment because of the nonlinear thermodynamics of refrigerants. Developing digital twin representatives for heat pumps capable of faster calculations with good accuracy is desirable. This study aimed to establish a generic deep learning–based digital twin for heat pumps with a large amount of high-fidelity data. Two refrigerants for two different heat pumps were considered: an air source heat pump with refrigerant R-410A, an air source heat pump with refrigerant CO 2 , a water source heat pump with refrigerant R-410A, and a water source heat pump with refrigerant CO 2 . Furthermore, results showed that the deep learning (long short-term memory) models effectively represented these four heat pumps as a digital twin: (a) accuracy for training and testing showed smaller than 0.02 for heating electricity and heating demands, and (b) the digital twins showed good consistency with original data for heating electricity and heating demands (root mean square errors of less than 0.12 W and 0.19 W, respectively). Therefore, deep learning–based heat pump models can be used in the co-simulation of building mechanical systems.

Air source heat pump↗
DOE OSTI2022

Global Barotropic Tide Modeling Using Inline Self‐Attraction and Loading in MPAS‐Ocean

Abstract We examine ocean tides in the barotropic version of the Model for Prediction Across Scales (MPAS‐Ocean), the ocean component of the Department of Energy Earth system model. We focus on four factors that affect tidal accuracy: self‐attraction and loading (SAL), model resolution, details of the underlying bathymetry, and parameterized topographic wave drag. The SAL term accounts for the tidal loading of Earth's crust and the self‐gravitation of the ocean and the load‐deformed Earth. A common method for calculating SAL is to decompose mass anomalies into their spherical harmonic constituents. Here, we compare a scalar SAL approximation versus an inline SAL using a fast spherical harmonic transform package. Wave drag accounts for energy lost by breaking internal tides that are produced by barotropic tidal flow over topographic features. We compare a series of successively finer quasi‐uniform resolution meshes (62.9, 31.5, 15.7, and 7.87 km) to a variable resolution (45 to 5 km) configuration. We ran MPAS‐Ocean in a single‐layer barotropic mode forced by five tidal constituents. The 45 to 5 km variable resolution mesh obtained the best total root‐mean‐square error (5.4 cm) for the deep ocean ( 1,000 m) tide compared to TPXO8 and ran twice as fast as the quasi‐uniform 8 km mesh, which had an error of 5.8 cm. This error is comparable to those found in other forward (non‐assimilative) ocean tide models. In future work, we plan to use MPAS‐Ocean to study tidal interactions with other Earth system components, and the tidal response to climate change.

54 ENVIRONMENTAL SCIENCES↗