SEARCH · Engineering Papers

Results for “ACCURACY”

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 91 records · Page 5

All sources NASA NTRS DOE OSTI

DOE OSTI2022

An initial investigation of accuracy required for the identification of small molecules in complex samples using quantum chemical calculated NMR chemical shifts

The majority of primary and secondary metabolites in nature have yet to be identified, representing a major challenge for metabolomics studies that currently require reference libraries from analyses of authentic compounds. Using currently available analytical methods, complete chemical characterization of metabolomes is infeasible for both technical and economic reasons. For example, unambiguous identification of metabolites is limited by the availability of authentic chemical standards, which, for the majority of molecules, do not exist. Computationally predicted or calculated data are a viable solution to expand the currently limited metabolite reference libraries, if such methods are shown to be sufficiently accurate. For example, determining nuclear magnetic resonance (NMR) spectroscopy spectra in silico has shown promise in the identification and delineation of metabolite structures. Many researchers have been taking advantage of density functional theory (DFT), a computationally inexpensive yet reputable method for the prediction of carbon and proton NMR spectra of metabolites. However, such methods are expected to have some error in predicted 13 >C and 1 H NMR spectra with respect to experimentally measured values. This leads us to the question–what accuracy is required in predicted 13 C and 1 H NMR chemical shifts for confident metabolite identification? Using the set of 11,716 small molecules found in the Human Metabolome Database (HMDB), we simulated both experimental and theoretical NMR chemical shift databases. We investigated the level of accuracy required for identification of metabolites in simulated pure and impure samples by matching predicted chemical shifts to experimental data. We found 90% or more of molecules in simulated pure samples can be successfully identified when errors of 1 H and 13 C chemical shifts in water are below 0.6 and 7.1 ppm, respectively, and below 0.5 and 4.6 ppm in chloroform solvation, respectively. In simulated complex mixtures, as the complexity of the mixture increased, greater accuracy of the calculated chemical shifts was required, as expected. However, if the number of molecules in the mixture is known, e.g., when NMR is combined with MS and sample complexity is low, the likelihood of confident molecular identification increased by 90%.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2024

Effects of machine learning errors on human decision-making: manipulations of model accuracy, error types, and error importance

Abstract This study addressed the cognitive impacts of providing correct and incorrect machine learning (ML) outputs in support of an object detection task. The study consisted of five experiments that manipulated the accuracy and importance of mock ML outputs. In each of the experiments, participants were given the T and L task with T-shaped targets and L-shaped distractors. They were tasked with categorizing each image as target present or target absent. In Experiment 1, they performed this task without the aid of ML outputs. In Experiments 2–5, they were shown images with bounding boxes, representing the output of an ML model. The outputs could be correct (hits and correct rejections), or they could be erroneous (false alarms and misses). Experiment 2 manipulated the overall accuracy of these mock ML outputs. Experiment 3 manipulated the proportion of different types of errors. Experiments 4 and 5 manipulated the importance of specific types of stimuli or model errors, as well as the framing of the task in terms of human or model performance. These experiments showed that model misses were consistently harder for participants to detect than model false alarms. In general, as the model’s performance increased, human performance increased as well, but in many cases the participants were more likely to overlook model errors when the model had high accuracy overall. Warning participants to be on the lookout for specific types of model errors had very little impact on their performance. Overall, our results emphasize the importance of considering human cognition when determining what level of model performance and types of model errors are acceptable for a given task.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2024

Moving toward automated µFTIR spectra matching for microplastic identification: addressing false identifications and improving accuracy

Abstract Infrared spectroscopy is a widely used tool for studying microplastics and identifying microparticles. Researchers rely on spectral libraries to differentiate between synthetic and natural materials. Unfortunately, spectral library matching is not perfect, and best practices require researchers to use time consuming, manual peak matching to assess spectral matches. Moving toward automated matching requires increased confidence in the matching process. Using spectra matching software may increase the efficiency of particle identification, however some matching strategies may confuse natural materials such as cotton, silk, and plant matter with common classes of synthetics such as polyesters and polyamides. In this experiment, we prepared 22 pristine sample materials from natural and synthetic sources and measured micro-Fourier transform infrared (µFTIR) spectra in transmission mode for each sample using a Thermo Nicolet iN10 MX instrument. The collected spectra were then input into two spectral library matching systems (Omnic Picta and Open Specy), using a total of five identification routines. Next, we placed a subset of four pristine microplastic materials in a biologically active river system for two weeks to simulate environmental samples. These simulated environmental samples were processed using 10% hydrogen peroxide for 24 h to remove organic contamination and then identified using the strongest performing library. We found that libraries with fewer sample spectra produced lower correlation matches and that using derivative correction greatly reduced the number of inaccuracies in identifying materials as either natural or synthetic. We also found that environmental fouling reduced the correlation value of library matches when compared to pristine particles, however the effect was not consistent across the four materials tested. Overall, we found that the accuracy of automated library matching in the tested systems and processing routines varied from 64.1 to 98.0% for distinguishing between natural and synthetic materials, and that a high Hit Quality Index (HQI) did not always correlate with accuracy. These results are important for the microplastic field, demonstrating a need to rigorously test spectral libraries and processing routines with known materials to ensure identification accuracy.

Kozloski, Rachel↗

DOE OSTI2023

Data and Scripts associated with: “Laboratory evaluation of open source and commercial electrical conductivity sensor precision and accuracy”

This data package is associated with the publication “Laboratory evaluation of open source and commercial electrical conductivity sensor precision and accuracy: How do they compare?” submitted to PLOS ONE and accepted for publication (Fulton, S.G. et al. 2023; doi not yet available).Variation in electrical conductivity (EC) of water is important to reveal environmental disturbance and natural dynamics, including factors such as anthropogenic salinization. This data package supports a study addressing the need for a robust performance assessment of open source (OS) EC sensors. We evaluated the accuracy (mean error, %) and precision (sample standard deviation) of OS EC sensors in the laboratory via comparison to EC calibration standards using three different OS and OS/commercial-hybrid (OS/C) EC sensors and data logger configurations and two commercial (C) EC sensors and data logger configurations. We also evaluated the effect of cable length (7.5 m and 30 m) and sensor calibration on OS sensor accuracy and precision. This data package presents the results from the different testing laboratory configurations. It also includes an R script for statistical analysis of the data and an Arduino IDE file used to calibrate and collect data with the OS Atlas EC sensor. File types are .csv, .pdf, .R, and .ino.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2024

Enhancing Fluid Flow Pressure and Saturation Prediction Accuracy and Reducing Uncertainty with Committee Machine – Illinois Basin Decatur Project (IBDP) as a Case Study

Presentation at the 17th International Conference on Greenhouse Gas Control Technologies GHGT-17 held in Calgary, Canada, October 20-24, 2024. Carbon capture and storage (CCS) is a way to play a critical role in the global transition to a low-emission economy. Current progress is hampered by a number of factors, among which the lack of risk-informed design tools and decision support frameworks is seen as a major roadblock. Significant interest exists in using artificial intelligence to accelerate CCS site feasibility studies, as well as to facilitate the permit application process. Existing works commonly train a single deep learning model. This work investigates the feasibility of using a conventional ensemble learning (committee machine) technique to further improve prediction accuracy. Ensemble-based algorithms generally improve over individual base learners in terms of robustness and accuracy. Deep ensembles, however, are time-consuming to create and train. A pragmatic question is whether small-sized ensembles may lead to prediction improvement. Here we evaluated the efficacy of an ensemble learning technique using the latent spectral model (LSM), an efficient deep neural operator algorithm, as base learners. Preliminary results, obtained using the Illinois Basin-Decatur Project (IBDP) carbon sequestration data/model, show that small-sized ensembles can improve prediction over the base learners, achieving prediction accuracy of ~1.6 psi root mean square error (RMSE) on pressure (relative the average reservoir pressure of 3150 psi), and less than 1.3% for saturation.

Sun, Alexander↗

DOE OSTI2026

Investigating Temperature Uniformity and Accuracy in PV Module Lamination: A Verification Study

This study investigates the temperature uniformity and accuracy of a photovoltaic (PV) module lamination process by addressing inconsistencies identified in 2017 data where irregular temperature changes were observed across setpoints. The 2017 data showed a notable drop in temperature upon bladder initiation, except for the 145 degrees Celsius profile. This inconsistency indicated potential inaccuracies in manual data recording methods. To address this concern, a verification experiment was conducted to evaluate temperature uniformity across the 2014 Bent River SPL2828 laminator platen and within test samples. Thermocouples, paired with Omega data acquisition software, were deployed to measure temperatures at multiple platen locations and within test samples. The experiment compared lamination temperatures of polyethylene-co-vinyl acetate (EVA) encapsulant when paired with solite glass or TPE backsheets. The methodology included verifying temperature uniformity directly on the platen and by using a large glass/EVA/glass sample using multiple thermocouples. Smaller samples were built with glass/EVA/glass and glass/EVA/backsheet configurations with one centered thermocouple to verify and compare sample temperatures. This verification aims to refine lamination temperature profiles, enhance data accuracy and provide insights into optimal process control for uniform module lamination. Ensuring consistent and uniform lamination may improve the accuracy and reliability of research outcomes.

14 SOLAR ENERGY↗

DOE OSTI2022

Analysis of Inlier and Outlier Compounds with Respect to Artificial Neural Network Cetane Number Prediction Accuracy

Artificial neural networks (ANNs) are exceptional at forming non-linear correlations between multivariate input and target variables; however, they are often seen as a “black box” approach, since how ANNs form these correlations is somewhat ambiguous. Furthermore, the process underlying how ANNs learn from inlier and outlier samples within the input dataset is not fully understood. Intuitively, it is expected that training ANNs with inlier samples will increase prediction accuracy and training with outlier samples will reduce prediction accuracy; though, in practice, this is not always true. The present work identifies and analyzes inliers and outliers of existing experimental cetane number (CN) data encompassing a variety of compounds and compound groups. It also investigates how ANNs trained to predict CN perform with and without outliers included in the training data, and whether a relationship exists between inliers/outliers and ANN prediction accuracy across the whole dataset and for individual samples. Additionally, individual outlier compounds are analyzed, highlighting how they structurally differ from inlier compounds.

09 BIOMASS FUELS↗

DOE OSTI2024

Accuracy optimized neural networks do not effectively model optic flow tuning in brain area MSTd

Accuracy-optimized convolutional neural networks (CNNs) have emerged as highly effective models at predicting neural responses in brain areas along the primate ventral stream, but it is largely unknown whether they effectively model neurons in the complementary primate dorsal stream. We explored how well CNNs model the optic flow tuning properties of neurons in dorsal area MSTd and we compared our results with the Non-Negative Matrix Factorization (NNMF) model, which successfully models many tuning properties of MSTd neurons. To better understand the role of computational properties in the NNMF model that give rise to optic flow tuning that resembles that of MSTd neurons, we created additional CNN model variants that implement key NNMF constraints – non-negative weights and sparse coding of optic flow. While the CNNs and NNMF models both accurately estimate the observer's self-motion from purely translational or rotational optic flow, NNMF and the CNNs with nonnegative weights yield substantially less accurate estimates than the other CNNs when tested on more complex optic flow that combines observer translation and rotation. Despite its poor accuracy, NNMF gives rise to tuning properties that align more closely with those observed in primate MSTd than any of the accuracy-optimized CNNs. This work offers a step toward a deeper understanding of the computational properties and constraints that describe the optic flow tuning of primate area MSTd.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2024

Reliability, biological variability, and accuracy of multi-frequency bioelectrical impedance analysis for measuring body composition components

Introduction Bioelectrical impedance analysis (BIA) systems are gaining popularity for use in research and fitness assessments as the technology improves and becomes more affordable and easier to use. Multifrequency BIA (MF-BIA) may improve accuracy and precision using octopolar contacts for segmental analyses. Purpose Evaluate reliability, biological variability, and accuracy of component measures (total body water, mass, and composition) of commercially available MF-BIA system (InBody 770, Cerritos, California, USA). Methods Fourteen healthy military-age adults were assessed by MF-BIA in duplicate on five laboratory visits across 3 weeks (10 measures each). Participants were evaluated at the same time of day after refraining from strenuous exercise (> 48 h), alcohol consumption (> 24 h), and caffeine, nicotine, and food (> 10 h). Systematic error (test–retest reliability) and biological variability (day-to-day reliability) were summarized by intraclass correlation coefficient (ICC) values determined for body mass (fat, fat-free, total) and body water (extracellular, intracellular, total). Body composition measurements derived from BIA on the second visit were also tested for accuracy compared to dual-energy x-ray absorptiometry (DXA). Results Test–retest reliability was very high for all measurements of whole-body water and mass (ICC ≥ 0.999) and high for regional body water and mass (ICC 0.973–1.000). Biological variability was observable with very minor differences between tests (same day) for total and regional body water (0.0–0.2 L) and total and regional body mass measurements (0.0–0.2 kg); while between day differences were slightly higher (0.0–0.5 L and 0.1–0.7 kg). Compared to DXA, the MF-BIA whole-body measures showed an offset in %BF (Bias −4.0 ± 2.8%; Standard error of the estimate (SEE), 2.6%), an overprediction for total body fat-free mass (Bias 2.8 ± 2.1 kg; SEE 2.2 kg) and an underprediction of total body fat mass (Bias −2.9 ± 2.0 kg; SEE 1.9 kg). Conclusion Under controlled conditions with fit and healthy men and women, this MF-BIA system has high methodological reliability and demonstrates stable day-to-day measurements of major body composition components. Previously reported ~3% body fat offset compared to criterion methods was again confirmed. Precision of the InBody 770 shows consistency and supports further testing of this specific device as a new military standards method and suitability across a wider range of %BF.

Nutrition & Dietetics↗

DOE OSTI2020

Multi-Trait Regressor Stacking Increased Genomic Prediction Accuracy of Sorghum Grain Composition

Genomic prediction has enabled plant breeders to estimate breeding values of unobserved genotypes and environments. The use of genomic prediction will be extremely valuable for compositional traits for which phenotyping is labor-intensive and destructive for most accurate results. We studied the potential of Bayesian multi-output regressor stacking (BMORS) model in improving prediction performance over single trait single environment (STSE) models using a grain sorghum diversity panel (GSDP) and a biparental recombinant inbred lines (RILs) population. A total of five highly correlated grain composition traits—amylose, fat, gross energy, protein and starch, with genomic heritability ranging from 0.24 to 0.59 in the GSDP and 0.69 to 0.83 in the RILs were studied. Average prediction accuracies from the STSE model were within a range of 0.4 to 0.6 for all traits across both populations except amylose (0.25) in the GSDP. Prediction accuracy for BMORS increased by 41% and 32% on average over STSE in the GSDP and RILs, respectively. Prediction of whole environments by training with remaining environments in BMORS resulted in moderate to high prediction accuracy. Our results show regression stacking methods such as BMORS have potential to accurately predict unobserved individuals and environments, and implementation of such models can accelerate genetic gain.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2021

A Data-Fusion Method using Bayesian Approach to Enhance Raw Data Accuracy of Position and Distance Measurements for Connected Vehicles

Accurate positioning of vehicles is a critical element of autonomous and connected vehicle systems. Most of other studies heavily focused on enhancing simultaneous localization and mapping (SLAM) methods, i.e., constructing or updating a map of an unknown environment and tracking an object within the map. This paper provides a method that can, in addition to existing SLAM or relevant methods, enhance the raw measurements of position and distance. The basic idea of this study is to identify and update the error distribution of each data source by combining all available information. A Bayesian approach was incorporated to estimate and update the error distribution of individual data sources or sensors. The proposed method can be conducted in real-time environments, and a self-learning scheme determines whether enough data has been collected to further improve the accuracy of such measurements. The simulated experiments show that the proposed model noticeably improves the accuracy of position and distance measurements. Especially, the estimated biases of position coordinates and distance measures are very close to the biases of true error distributions, with the R-squared over 0.98. A similar approach can also be utilized to enhance accuracy of other sensors or measurements in connected vehicle or relevant systems, where multi-data sources are available.

Lim, Hyeonsup↗

DOE OSTI2024

Enhancing Fluid Flow Pressure and Saturation Prediction Accuracy and Reducing Uncertainty with Committee Machine – Illinois Basin Decatur Project (IBDP) as a Case Study

This is the conference paper accompanying an oral presentation at the 17th International Conference on Greenhouse Gas Control Technologies GHGT-17 held in Calgary, Canada, October 20-24, 2024. Carbon capture and storage (CCS) is a way to play a critical role in the global transition to a low-emission economy. Current progress is hampered by a number of factors, among which the lack of risk-informed design tools and decision support frameworks is seen as a major roadblock. Significant interest exists in using artificial intelligence to accelerate CCS site feasibility studies, as well as to facilitate the permit application process. Existing works commonly train a single deep learning model. This work investigates the feasibility of using a conventional ensemble learning (committee machine) technique to further improve prediction accuracy. Ensemble-based algorithms generally improve over individual base learners in terms of robustness and accuracy. Deep ensembles, however, are time-consuming to create and train. A pragmatic question is whether small-sized ensembles may lead to prediction improvement. Here we evaluated the efficacy of an ensemble learning technique using the latent spectral model (LSM), an efficient deep neural operator algorithm, as base learners. Preliminary results, obtained using the Illinois Basin-Decatur Project (IBDP) carbon sequestration data/model, show that small-sized ensembles can improve prediction over the base learners, achieving prediction accuracy of ~1.6 psi root mean square error (RMSE) on pressure (relative the average reservoir pressure of 3150 psi), and less than 1.3% for saturation.

Sun, Alexander↗

DOE OSTI2021

Shake loss intensities in x-ray photoelectron spectroscopy: Theory, experiment, and atomic composition accuracy for MgO and related compounds

The relative intensities of XPS core levels, scaled by their photoionization cross sections, are regularly used to determine sample atomic composition. Cross sections, however, give the intensity to all possible final states for the core ionizations, not just to the main peak. This includes all intrinsic satellite structure (shake states and, for open shell systems, the different ionic multiplets). In practice, for solids, this is usually experimentally impossible to determine accurately because such a satellite structure sits on the inelastically scattered electron background and cannot be easily separated. Therefore, usually, only the intensity of the main peak is used. This limits the ultimate possible accuracy of XPS composition determination. The purpose of the present paper is to examine the contributions that a theoretical analysis of losses of intensity can make to improve quantitation. For an MgO single crystal, we show that the correct stoichiometry of 1:1 can be recovered using the theoretical analysis of the experimental MgO peak ratio intensities. For materials with a sufficient bandgap for the XPS main peaks to be separated from the scattered background, the intensity of main peaks can often be accurately determined. Thus, if one uses theory to calculate that fraction of the total intensity lost from a main peak into all its satellite structure, the intensity of just main peaks could then be used to more accurately determine relative atom % composition. This work tests this approach using a single crystal MgO (50% Mg, 50% O) standard. Ab initio electronic structure theory of representative MgO clusters is used to determine Hartree–Fock wave functions for the ground state and final ionized states corresponding to the main Mg 2p and O1s XPS peaks of the oxide. The sudden approximation, SA, is used to determine the fractional losses from these main peaks to shake satellites, which is found to be greater for O1s than Mg2p. This results in predicted “apparent composition” for stoichiometric MgO of 55.2% Mg, 44.8% O instead of the true 50% Mg, 50% O. Equivalent theory for CaO results in a predicted apparent Ca value of 53.4%. Experimentally, using Mg2s or 2p intensity ratio to O1s, we find values between 52.2% and 56.0% Mg using two crystals and four different instrument electron pass energies. The average value of the measurements is 54.5% Mg when corrected for the presence of an adventitious carbon overlayer and slight surface hydroxide. Though this agreement with theory may be somewhat fortuitous, given the potential experimental errors, which are fully discussed, it is similar to that in our earlier study on LiF. We also present preliminary experimental data on Mg(OH) 2 and MgSO 4 , which show a similar trend of apparently higher than 50% Mg, but we have no theory values. We are not yet able to experimentally test for validation of the difference between apparent composition for MgO (55.2% Mg) and CaO (53.4% Ca), owing to significant carbonate formation at the surface of the single crystal CaO. Here, an important conclusion is that the theoretical determination of shake losses, obtained with ab initio wavefunctions and the SA, is likely to be a useful way to calibrate the accuracy and reliability of compositions obtained from XPS intensities and merits further study.

47 OTHER INSTRUMENTATION↗

DOE OSTI2023

Best Practices for Body Temperature Measurement with Infrared Thermography: External Factors Affecting Accuracy

Infrared thermographs (IRTs) are commonly used during disease pandemics to screen individuals with elevated body temperature (EBT). To address the limited research on external factors affecting IRT accuracy, we conducted benchtop measurements and computer simulations with two IRTs, with or without an external temperature reference source (ETRS) for temperature compensation. The combination of an IRT and an ETRS forms a screening thermograph (ST). We investigated the effects of viewing angle (θ, 0–75°), ETRS set temperature (T ETRS , 30–40 °C), ambient temperature (T atm , 18–32 °C), relative humidity (RH, 15–80%), and working distance (d, 0.4–2.8 m). We discovered that STs exhibited higher accuracy compared to IRTs alone. Across the tested ranges of T atm and RH, both IRTs exhibited absolute measurement errors of less than 0.97 °C, while both STs maintained absolute measurement errors of less than 0.12 °C. The optimal T ETRS for EBT detection was 36–37 °C. When θ was below 30°, the two STs underestimated calibration source (CS) temperature (T CS ) of less than 0.05 °C. The computer simulations showed absolute temperature differences of up to 0.28 °C and 0.04 °C between estimated and theoretical temperatures for IRTs and STs, respectively, considering d of 0.2–3.0 m, T atm of 15–35 °C, and RH of 5–95%. The results highlight the importance of precise calibration and environmental control for reliable temperature readings and suggest proper ranges for these factors, aiming to enhance current standard documents and best practice guidelines. These insights enhance our understanding of IRT performance and their sensitivity to various factors, thereby facilitating the development of best practices for accurate EBT measurement.

47 OTHER INSTRUMENTATION↗

DOE OSTI2024

Machine learning for accuracy in density functional approximations

Machine learning techniques have found their way into computational chemistry as indispensable tools to accelerate atomistic simulations and materials design. In addition, machine learning approaches hold the potential to boost the predictive power of computationally efficient electronic structure methods, such as density functional theory, to chemical accuracy and to correct for fundamental errors in density functional approaches. In this paper, recent progress in applying machine learning to improve the accuracy of density functional and related approximations is reviewed. Promises and challenges in devising machine learning models transferable between different chemistries and materials classes are discussed with the help of examples applying promising models to systems far outside their training sets.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2024

Evaluating Cryo–TEM Reconstruction Accuracy of Self–Assembled Polymer Nanostructures

Cryogenic transmission electron microscopy (cryo–TEM) combined with single particle analysis (SPA) is an emerging imaging approach for soft materials. However, the accuracy of SPA–reconstructed nanostructures, particularly those formed by synthetic polymers, remains uncertain due to potential packing heterogeneity of the nanostructures. In this study, the combination of molecular dynamics (MD) simulations and image simulations is utilized to validate the accuracy of cryo–TEM 3D reconstructions of self–assembled polypeptoid fibril nanostructures. Using CryoSPARC software, image simulations, 2D classifications, ab initio reconstructions, and homogenous refinements are performed. By comparing the results with atomic models, the recovery of molecular details is assessed, heterogeneous structures are identified, and the influence of extraction location on the reconstructions is evaluated. In conclusion, these findings confirm the fidelity of single particle analysis in accurately resolving complex structural characteristics and heterogeneous structures, exhibiting its potential as a valuable tool for detailed structural analysis of synthetic polymers and soft materials.

36 MATERIALS SCIENCE↗

DOE OSTI2022

Empirical relationships between environmental factors and soil organic carbon produce comparable prediction accuracy as the Machine Learning

Accurate representation of environmental controllers of soil organic carbon (SOC) stocks in Earth System Model (ESM) land models could reduce uncertainties in future carbon-climate feedback projections. Using empirical relationships between environmental factors and SOC stocks to evaluate land models can help modelers understand prediction biases beyond what can be achieved with the observed SOC stocks alone. In this study, we used 31 observed environmental factors, field SOC observations (n = 6,213) from the continental US, and two Machine Learning approaches [Random Forest (RF) and Generalized Additive Modeling (GAM)] to (1) select important environmental predictors of SOC stocks, (2) derive empirical relationships between environmental factors and SOC stocks, and (3) use the derived relationships to predict SOC stocks and compare the prediction accuracy of simpler model developed with the machine learning predictions. Out of the 31 environmental factors we investigated, 12 were identified as important predictors of SOC stocks by the RF approach. In contrast, the GAM approach identified six (of those 12) environmental factors as important controllers of SOC stocks: potential evapotranspiration, normalized difference vegetation index, soil drainage condition, precipitation, elevation, and net primary productivity. The GAM approach showed minimal SOC predictive importance of the remaining six environmental factors identified by the RF approach. Our derived empirical relations produced comparable prediction accuracy as the GAM and RF approach using only a subset of environmental factors. The empirical relationships we derived using the GAM approach can serve as important benchmarks to evaluate environmental control representations of SOC stocks in ESMs, which could reduce uncertainty in predicting future carbon-climate feedbacks.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2024

Quantifying Epistemic Uncertainty in Binary Classification via Accuracy Gain

ABSTRACT Recently, a surge of interest has been given to quantifying epistemic uncertainty (EU), the reducible portion of uncertainty due to lack of data. We propose a novel EU estimator in the binary classification setting, as the posterior expected value of the empirical gain in accuracy between the current prediction and the optimal prediction. In order to validate the performance of our EU estimator, we introduce an experimental procedure where we take an existing dataset, remove a set of points, and compare the estimated EU with the observed change in accuracy. Through real and simulated data experiments, we demonstrate the effectiveness of our proposed EU estimator.

97 MATHEMATICS AND COMPUTING↗