Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Scientific method”

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 523 records · Page 29
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Phage display and other peptide display technologies

ABSTRACT Phage display technology, which is based on the presentation of peptide sequences on the surface of bacteriophage virions, was developed over 30 years ago. Improvements in phage display systems have allowed us to employ this method in numerous fields of biotechnology, as diverse as immunological and biomedical applications, the formation of novel materials and many others. The importance of phage display platforms was recognized by awarding the Nobel Prize in 2018 ‘for the phage display of peptides and antibodies’. In contrast to many review articles concerning specific applications of phage display systems published in recent years, we present an overview of this technology, including a comparison of various display systems, their advantages and disadvantages, and examples of applications in various fields of science, medicine and the broad sense of biotechnology. Other peptide display technologies, which employ bacterial, yeast and mammalian cells, as well as eukaryotic viruses and cell-free systems, are also discussed. These powerful methods are still being developed and improved; thus, novel sophisticated tools based on phage display and other peptide display systems are constantly emerging, and new opportunities to solve various scientific, medical and technological problems can be expected to become available in the near future.

Jaroszewicz, Weronika (ORCID:0000000302948682)↗
DOE OSTI2026

Block encoding of the three-dimensional heterogeneous Poisson equation with application to fracture flow

Quantum linear system (QLS) algorithms offer the potential to solve large-scale linear systems exponentially faster than classical methods. However, applying QLS algorithms to real-world problems remains challenging due to issues such as state preparation, data loading, and efficient information extraction. In this work, we study the feasibility of applying QLS algorithms to solve discretized three-dimensional (3D) heterogeneous Poisson equations, with specific examples relating to groundwater flow through geologic fracture networks. We explicitly construct a block encoding for the 3D heterogeneous Poisson matrix by leveraging the sparse local structure of the discretized operator. While classical solvers benefit from preconditioning, we show that block encoding the system matrix and preconditioner separately does not improve the effective condition number that dominates the QLS run-time. This differs from classical approaches where the preconditioner and the system matrix can often be implemented independently. Nevertheless, due to the structure of the problem in three dimensions, the quantum algorithm achieves a run-time of 𝑂⁡(𝑁 2/3 polylog 𝑁 ⋅log (1/𝜖)), outperforming the best classical methods (with run times of 𝑂⁡(𝑁⁢log 𝑁 ⋅log (1/𝜖))) and offering exponential memory savings. These results highlight both the promise and limitations of QLS algorithms for practical scientific computing, and point to effective condition-number reduction as a key barrier in achieving quantum advantages.

58 GEOSCIENCES↗
DOE OSTI2020

Quantifying redox heterogeneity in single-crystalline LiCoO 2 cathode particles

Active cathode particles are fundamental architectural units for the composite electrode of Li-ion batteries. The microstructure of the particles has a profound impact on their behavior and, consequently, on the cell-level electrochemical performance. LiCoO 2 (LCO, a dominant cathode material) is often in the form of well-shaped particles, a few micrometres in size, with good crystallinity. In contrast to secondary particles (an agglomeration of many fine primary grains), which are the other common form of battery particles populated with structural and chemical defects, it is often anticipated that good particle crystallinity leads to superior mechanical robustness and suppressed charge heterogeneity. Yet, sub-particle level charge inhomogeneity in LCO particles has been widely reported in the literature, posing a frontier challenge in this field. Herein, this topic is revisited and it is demonstrated that X-ray absorption spectra on single-crystalline particles with highly anisotropic lattice structures are sensitive to the polarization configuration of the incident X-rays, causing some degree of ambiguity in analyzing the local spectroscopic fingerprint. To tackle this issue, a methodology is developed that extracts the white-line peak energy in the X-ray absorption near-edge structure spectra as a key data attribute for representing the local state of charge in the LCO crystal. This method demonstrates significantly improved accuracy and reveals the mesoscale chemical complexity in LCO particles with better fidelity. In addition to the implications on the importance of particle engineering for LCO cathodes, the method developed herein also has significant impact on spectro-microscopic studies of single-crystalline materials at synchrotron facilities, which is broadly applicable to a wide range of scientific disciplines well beyond battery research.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2021

An Investigation into High Voltage Spiral Generators Utilizing Thyristor Input Switches

High voltage nanosecond pulses are widely used in scientific research, but their wider adoption in industry requires compact, cost effective and easy to use generators to be developed. This paper presents modelling and experimental investigations into one method of producing such pulses – a spiral generator with a solid-state, thyristor-based input switch. It includes how the pulses are formed within the spiral, why a high speed input switch is required, how the geometry of the spiral dictates its output characteristics, and the effects of different loads. Using thyristors, often connected in series to increase the operating voltage of the spiral, enables the spiral generators to have low jitter, high repetition rate, and long lifetime. Modelling of the circuit used a combination of telegraph equations to account for the wave propagation along the spiral and a lumped circuit exchanging charge between the spiral, the input switch, and the load. The model is verified by detailed experimental results with the relative error being < 10% in most cases. Here, the output voltage pulse was often observed to have an initial peak of much lower magnitude than the subsequent peak(s) – which can only be fully explained by considering wave propagation effects. Lower input switch inductance, shorter switching time, larger mean diameter of the spiral, and increasing the width of the copper tape that makes up the spiral can all increase the voltage multiplication efficiency. Though increasing the number of turns that make up the spiral can increase the output voltage, it can also lower the multiplication efficiency. By understanding the effects of different geometries the spiral can be optimized to drive different loads - three applications of such spiral generators are then presented: pulses with 10 kV amplitude and 10 kHz repetition rate for driving DBD plasma, pulses with amplitude of 10 kV and 10 kV/ns rising rate for triggering of advanced solid state switches, and pulses with -50 kV amplitude and 50 ns rising time for triggering high current gas switches through field distortion.1 Index Terms—spiral generator, nanosecond pulses, telegraph equations, series-connected thyristor module, pulse forming process

42 ENGINEERING↗
DOE OSTI2021

BETO 2021 Peer Review - Analytical Development & Support WBS 2.5.1.101

The objective of the Analytical Development and Support (ADS) Project is to produce and maintain the critical analytical methods and tools that enable evaluation of emerging biofuels R&D at NREL and in the broader biofuels research community. Our project is divided into two tasks: one task to develop novel analytical techniques and improve existing methods and one task to maintain existing analytical capabilities at NREL and provide outreach to the wider community. The ADS Project is world-recognized for our Laboratory Analytical Procedures (LAPS) which provide detailed procedures for compositional analysis of biomass and have been adopted as the de facto standards within the biofuels community largely due to the transparency of the methods and the high reputation of NREL's research. Our dialog with stakeholders allows us to provide robust, precise, accurate, and publicly available analytical procedures for better valuation of scientific tools such as our recent accomplishment in developing a cellulose assay to support the EPA and industry in calculation of converted cellulose during starch ethanol production. We continue to develop analytical capabilities to support BETO's directives to research cost advantaged fuels such as animal wastes and novel bioproducts like 2,3-Butane-diol.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2020

2020 Exascale Computing Project Annual Meeting (Executive Summary Report)

The Exascale Computing Project (ECP) delivers specific applications, software products, and outcomes on DOE computing facilities. Integration across these elements for specific hardware technologies for exascale system instantiations is fundamental to ECP success. The outcome of the ECP is the delivery of a capable exascale computing ecosystem to provide breakthrough solutions addressing our most critical challenges in scientific discovery, energy assurance, economic competitiveness, and national security. This outcome is not a matter of ensuring more powerful computing systems. The ECP is designed to create more valuable and rapid insights from a wide variety of applications (“capable”), which requires a much higher level of inherent efficacy in all methods, software tools, and ECP-enabled computing technologies to be acquired by DOE laboratories (“ecosystem”). The ECP annual meeting provides a unique opportunity for the core technical expertise in the United States focused on achieving this next plateau of computational science and computing performance to engage in direct discussions on project execution. Face-to-face gatherings in technical communities like this are common and needed for the exchange of scientific ideas and technical performance. The ECP annual meeting stands apart from other technical conferences and meetings in the computing community as it is uniquely and solely focused on the execution of the ECP and the integration of technical activities leading to the creation of the exascale computing ecosystem for the future. The direct interaction of key critical technical staff, who are leaders in their respective fields, and the resulting give-and-take between software, applications, and hardware and the technical co-design therein, is unique and essential to the effective execution of the ECP. The first annual meeting was held in Knoxville, Tennessee, January 31 – February 2, 2017 and brought together, for the first time, a diverse collection of researchers from 16 DOE national laboratories as well as university computer and computational science researchers to discuss shared problems and joint solutions for the development of a capable exascale computing ecosystem. These interactions resulted in focused technical plans and an energized community centered on advances for ECP. The second annual meeting was held in Knoxville, Tennessee, February 5–9, 2018. It included 643 individual thought leaders and performers in application development, software research and deployment, and hardware research and integrators, all of whom are part of the multifaceted, billion dollar HPC community. This meeting provided a platform to discuss and disseminate numerous examples where researchers with common goals and synergistic solutions came together for the first time to deliver tangible results. Additionally, at the 2018 meeting, ECP researchers had the opportunity to digest all US HPC vendor R&D product roadmaps pointing to exascale – not only to learn how their research can play a role, but, more importantly, to influence those roadmaps to ensure successful delivery on DOE applications that will contribute to (if not solve) problems of national interest in national security, science, energy, and health, as well as growing security threats. The third annual meeting was held in Houston, Texas, January 14–17, 2019. With a 19% increase in the number of registrations (768 people), and the change in location, the third annual meeting was considered the most impactful of the three at the time. The new website provided a better platform for the dissemination of the content, the new venue as a meeting hotel instead of a conference center facilitated interactions and discussions after event hours, and the addition of an award-winning mobile event conference app (Whova) transformed dramatically the attendee experience at the event. This fourth annual meeting was held in Houston, Texas, February 3-7, 2020. This meeting had an increase in the number of attendees for a total of 824 people registered (782 attendees) and included numerous enhancements based on feedback and lessons learned from previous meetings, some of which are listed here: Improved quality of the sessions, their material and the whole program.; Had more industry participation and addition of external collaborators from overseas.; Published the full agenda earlier to better accommodate attendance and travel plans based on schedule.; Centralized all sessions in one venue.; Provided additional hotels and room blocks for the attendees.; Improved communication with the audience (links, material, directions, notifications, etc.) to go paperless.; Enhanced side meeting scheduling, management and user experience.; Made available additional space and tables for impromptu meetings and side discussions.; Improved IT and A/V solutions for speakers. In addition, our final survey captured the following points as opportunities for improvement in future meetings: consider a different meeting location that is more pedestrian friendly, reduce talks during working meals to allow more collaboration and informal time, adapt the agenda to acknowledge attendees from different timezones, consider recording some of the tutorials and/or sessions to share broadly with the HPC community, have a larger poster room, provide additional power strips, and improve the WiFi.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

A General Framework for Progressive Data Compression and Retrieval

In scientific simulations, observations, and experiments, the transfer of data to and from disk and across networks has become a major bottleneck for data analysis and visualization. Compression techniques have been employed to tackle this challenge, but traditional lossy methods often demand conservative error tolerances to meet the numerical accuracy requirements of both anticipated and unknown data analysis tasks. Progressive data compression and retrieval has emerged as a promising solution, where each analysis task dictates its own accuracy needs. However, few analysis algorithms inherently support progressive data processing, and adapting compression techniques, file formats, client/server frameworks, and APIs to support progressivity can be challenging. Here, this paper presents a framework that enables progressive-precision data queries for any data compressor or numerical representation. Our strategy hinges on a multi-component representation that successively reduces the error between the original and compressed field, allowing each field in the progressive sequence to be expressed as a partial sum of components. We have implemented this approach with four established scientific data compressors and assessed its effectiveness using real-world data sets from the SDRBench collection. The results show that our framework competes in accuracy with the standalone compressors it is based upon. Additionally, (de)compression time is proportional to the number of components requested by the user. Finally, our framework allows for fully lossless compression using lossy compressors when a sufficient number of components are employed.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2020

Transuranic Waste Volume Reduction Methods for Los Alamos National Laboratory

Los Alamos National Laboratory (LANL, or the Lab) is one of the United States’ (U.S.) weapons complex nuclear security laboratories; it is owned by the Department of Energy (DOE) and operated by Triad National Security, LLC. LANL’s mission is to solve national security challenges through scientific excellence, which includes nuclear deterrence and stockpile stewardship. Such activities generate a variety of radioactive waste types, including transuranic (TRU) waste. The National Nuclear Security Administration (NNSA) is under a congressional mandate to produce no fewer than 80 plutonium pits per year by 2030 to meet Department of Defense requirements and support the nuclear deterrent. LANL is responsible for producing at least 30 pits per year by 2026, while the Savannah River Site will produce 50 pits per year by 2030. As of August 2018, the NNSA TRU waste inventory at LANL was at 53% capacity. In order to ensure the nuclear material missions are able to continue uninterrupted, the storage of TRU waste needs to remain below 60% of total capacity. As LANL increases production and prepares to meet the 30 pits per year goal, waste generation will increase substantially. Thus, there is a need to rejuvenate current infrastructure and implement TRU waste reduction methods in the coming years.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2026

Opportunities in AI/ML for the Rubin LSST Dark Energy Science Collaboration

The Vera C. Rubin Observatory's Legacy Survey of Space and Time (LSST) will produce unprecedented volumes of heterogeneous astronomical data (images, catalogs, and alerts) that challenge traditional analysis pipelines. The LSST Dark Energy Science Collaboration (DESC) aims to derive robust constraints on dark energy and dark matter from these data, requiring methods that are statistically powerful, scalable, and operationally reliable. Artificial intelligence and machine learning (AI/ML) are already embedded across DESC science workflows, from photometric redshifts and transient classification to weak lensing inference and cosmological simulations. Yet their utility for precision cosmology hinges on trustworthy uncertainty quantification, robustness to covariate shift and model misspecification, and reproducible integration within scientific pipelines. This white paper surveys the current landscape of AI/ML across DESC's primary cosmological probes and cross-cutting analyses, revealing that the same core methodologies and fundamental challenges recur across disparate science cases. Since progress on these cross-cutting challenges would benefit multiple probes simultaneously, we identify key methodological research priorities, including Bayesian inference at scale, physics-informed methods, validation frameworks, and active learning for discovery. With an eye on emerging techniques, we also explore the potential of the latest foundation model methodologies and LLM-driven agentic AI systems to reshape DESC workflows, provided their deployment is coupled with rigorous evaluation and governance. Finally, we discuss critical software, computing, data infrastructure, and human capital requirements for the successful deployment of these new methodologies, and consider associated risks and opportunities for broader coordination with external actors.

Aubourg, Eric [APC, Paris] (ORCID:000000025592023X↗
DOE OSTI2024

Acetaminophen production in the edible, filamentous cyanobacterium Arthrospira platensis

Abstract Spirulina is the common name for the edible, nonheterocystous, filamentous cyanobacteriumArthrospira platensisthat is grown industrially as a food supplement, animal feedstock, and pigment source. Although there are many applications for engineering this organism, until recently no genetic tools or reproducible transformation methods have been published. While recent work showed the production of a diversity of proteins inA. platensis, including single‐domain antibodies for oral delivery, there remains a need for a modular, characterized genetic toolkit. Here, we independently establish a reproducible method for the transformation ofA. platensisand engineer this bacterium to produce acetaminophen as proof‐of‐concept for small molecule production in an edible host. This work opensA. platensisto the wider scientific community for future engineering as a functional food for nutritional enhancement, modification of organoleptic traits, and production of pharmaceuticals for oral delivery.

Biotechnology & Applied Microbiology↗
DOE OSTI2020

Precision inclusive Higgs physics at e+e- colliders with tracking detectors and without calorimetry

A primary goal of a future e + e - collider program will be the precision measurement of Higgs boson properties. For practical reasons it is of interest to determine the minimal set of detector specifications required to reach this and other scientific goals. Here we investigate the precision obtainable for the e + e - Zh μ + μ - X inclusive cross section and the Higgs boson mass using the di-muon recoil method, considering a detector that has only an inner tracking system within a solenoidal magnetic field, surrounded by many nuclear interaction lengths of absorbing material, and an outer muon identification system. We find that the sensitivity achievable in these measurements with such a tracking detector is only marginally reduced compared to that expected for a general purpose detector with additional electromagnetic and hadronic calorimeter systems. The difference results mainly from multi-photon backgrounds that are not as easily rejected with tracking detectors. We also comment on the prospects for an analogous measurement of the e + e - → Zh → e + e - X inclusive cross section. Finally, we study searches for light scalars utilizing the di-muon recoil method, estimating the projected reach with a tracking or general purpose detector.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2024

CG-Kit: Code Generation Toolkit for performant and maintainable variants of source code applied to Flash-X hydrodynamics simulations

CG-Kit is a new Code Generation tool-Kit that we have developed as a part of the solution for portability and maintainability for multiphysics computing applications. The development of CG-Kit is rooted in the urgent need created by the shifting landscape of high-performance computing platforms and the algorithmic complexities of a particular large-scale multiphysics application: Flash-X. To efficiently use computing resources on a heterogeneous node, an application must have a map of computation to resources and a mechanism to move the data and computation to the resources according to the map. Most existing performance portability solutions are focussed on abstracting the expression of computations so that a unified source code can be specialized to run on different resources. However, such an approach is insufficient for a code like Flash-X, which has a multitude of code components that can be assembled in various permutations and combinations to form different instances of applications. Similar challenges apply to any code that has composability, where a single specified way of apportioning work among devices may not be optimal. Additionally, use cases arise where the optimal control flow of computation may differ for different devices while the underlying numerics remain identical. This combination leads to unique challenges including handling an existing large code base in Fortran and/or C/C++, subdivision of code into a great variety of units supporting a wide range of physics and numerical methods, different parallelization techniques for distributed and shared memory systems and accelerator devices, and heterogeneity of computing platforms requiring coexisting variants of parallel algorithms. All of these challenges demand that scientific software developers apply existing knowledge about domain applications, algorithms, and computing platforms to determine custom abstractions and granularity for code generation. There is a critical lack of tools to tackle those problems. CG-Kit is designed to fill this gap by providing a user with the ability to express their desired control flow and computation-to-resource map in the form a pseudocode-like recipe. It consists of standalone tools that can be combined into highly specific and, we argue, highly effective portability and maintainability toolchains. Here we present the design of our new tools: parametrized source trees, control flow graphs, and recipes. The tools are implemented in Python. They are agnostic to the programming language of the source code targeted for code generation. In conclusion, we demonstrate the capabilities of the toolkit with two examples, first, multithreaded variants of the basic AXPY operation, and second, variants of parallel algorithms within a hydrodynamics solver, called Spark, from Flash-X that operates on block-structured adaptive meshes.

Algorithmic portability↗
DOE OSTI2021

Engineering the beat phenomenon of quasi-Rayleigh waves for regions with minimal Surface Acoustic Wave (SAW) amplitude

Surface Acoustic Waves (SAW) are prevalent in a variety of scientific fields. Typically, the ability to manipulate these surface waves has a significant impact on the efficacy of any SAW-based device. In this paper, we present a simple method for designing localized regions that have minimal SAW amplitude. These localized regions have valley-like drops in SAW amplitude and are achieved by using the spatial beat phenomenon of quasi-Rayleigh waves. Controlling the beat phenomenon allows us to manipulate where and how these localized regions appear in the waveguide. In particular, the localized regions can be designed to have variable widths and can be positioned consecutively due to minimal wave scattering. The design methodology and physics are described in detail, and several supplementary simulations are provided to demonstrate the efficacy of the proposed method. Finally, we briefly discuss how the proposed approach could potentially be exploited to implement a localized high-pass filter and enables SAW wave propagation that avoids surface obstacles.

36 MATERIALS SCIENCE↗
DOE OSTI2026

Bayesian Optimization of Catalysis with In-Context Learning

Large language models (LLMs) can perform accurate classification with zero or few examples through in-context learning (ICL), allowing the model to observe query-relevant examples at inference time and eliminating the need for additional weight updates to generalize beyond its original training data. We extend this capability to regression with uncertainty estimation using frozen LLMs (e.g., GPT-4o, Gemini), enabling Bayesian optimization (BO) in natural language without explicit model training or feature engineering. We apply this to materials discovery by representing materials as synthesis and testing procedures for use in natural language prompts. This Bayesian, design-first approach prioritizes optimization toward target material properties before detailed characterization, in contrast to conventional experimental workflows that often emphasize characterization of suboptimal materials. On benchmarks like aqueous solubility and oxidative coupling of methane (OCM), BO-ICL matches or outperforms Gaussian processes. In live experiments on the reverse water–gas shift (RWGS) reaction, BO-ICL identifies multimetallic catalysts that approach equilibrium CO yield within 6 and 10 iterations from a pool of 3,700 and 360,000 candidates, respectively. Our method redefines materials representation and accelerates discovery, with broad applications across catalysis, materials science, and AI.

Calibration↗
DOE OSTI2023

Deep learning at the edge enables real-time streaming ptychographic imaging

Abstract Coherent imaging techniques provide an unparalleled multi-scale view of materials across scientific and technological fields, from structural materials to quantum devices, from integrated circuits to biological cells. Driven by the construction of brighter sources and high-rate detectors, coherent imaging methods like ptychography are poised to revolutionize nanoscale materials characterization. However, these advancements are accompanied by significant increase in data and compute needs, which precludes real-time imaging, feedback and decision-making capabilities with conventional approaches. Here, we demonstrate a workflow that leverages artificial intelligence at the edge and high-performance computing to enable real-time inversion on X-ray ptychography data streamed directly from a detector at up to 2 kHz. The proposed AI-enabled workflow eliminates the oversampling constraints, allowing low-dose imaging using orders of magnitude less data than required by traditional methods.

36 MATERIALS SCIENCE↗
DOE OSTI2020

Ozone Gas: Scientific Justification and Practical Guidelines for Improvised Disinfection using Consumer-Grade Ozone Generators and Plastic Storage Boxes

Here, we describe direct measurements of ozone concentration achievable in small enclosed containers (plastic storage boxes) for use as improvised decontamination systems for small articles such as disposable PPE (N95 masks, nitrile gloves, etc.), clothing, mail and small packages, food, and other miscellaneous articles. The emphasis is on the reliable and sustained generation of ozone gas concentrations of sufficient concentration and duration to create an effective virucidal environment to achieve more than 95% to 99% viral inactivation, based upon the data already published in the peer-review literature on this topic. The suggestion that ozone be used to inactivate virus is certainly not a new idea. Our objective in this report is to make clear that the necessary levels of ozone can be improvised using simple, easy-to-use, inexpensive, and widely available supplies, and that there is every theoretical and experimental reason to believe that this approach is as highly effective in viral inactivation by ozone as are the far more expensive, complex, cumbersome, and less available equivalent ozone (and other) disinfectant systems that have themselves become unavailable during times of pandemic crisis. Using multiple types of readily available commercial ozone generators, concentration in the tested improvised enclosure is tracked over time to assess ozone charging and decay rates, and the ozone quenching effects of items placed in the box. Generator performance is compared against published ozone dosage values for virucidal and antimicrobial activity. Bubbler and box-fan-type ozone generators were found to be effective at achieving and maintaining target concentrations of 10ppm ozone or higher, whereas automotive cigarette lighter and universal serial bus type plug in “air freshener” ozone generators could not achieve the target concentrations in these experiments. Calculations and practical guidelines for assembly and effective use of an ozone box for improvised decontamination are offered. The majority of this report is directed toward the scientific justification and rationale for this approach. The end of the document summarizes the findings and offers simplified designs for the construction and use of ozone boxes as an improvised method of disinfection.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2021

Development of observation-based global multilayer soil moisture products for 1970 to 2016

Abstract. Soil moisture (SM) datasets are critical to understanding the global water, energy, and biogeochemical cycles and benefit extensive societal applications. However, individual sources of SM data (e.g., in situ and satellite observations, reanalysis, offline land surface model simulations, Earth system model – ESM – simulations) have source-specific limitations and biases related to the spatiotemporal continuity, resolutions, and modeling and retrieval assumptions. Here, we developed seven global, gap-free, long-term (1970–2016), multilayer (0–10, 10–30, 30–50, and 50–100 cm) SM products at monthly 0.5∘ resolution (available at https://doi.org/10.6084/m9.figshare.13661312.v1; Wang and Mao, 2021) by synthesizing a wide range of SM datasets using three statistical methods (unweighted averaging, optimal linear combination, and emergent constraint). The merged products outperformed their source datasets when evaluated with in situ observations (mean bias from −0.044 to 0.033 m3 m−3, root mean square errors from 0.076 to 0.104 m3 m−3, Pearson correlations from 0.35 to 0.67) and multiple gridded datasets that did not enter merging because of insufficient spatial, temporal, or soil layer coverage. Three of the new SM products, which were produced by applying any of the three merging methods to the source datasets excluding the ESMs, had lower bias and root mean square errors and higher correlations than the ESM-dependent merged products. The ESM-independent products also showed a better ability to capture historical large-scale drought events than the ESM-dependent products. The merged products generally showed reasonable temporal homogeneity and physically plausible global sensitivities to observed meteorological factors, except that the ESM-dependent products underestimated the low-frequency temporal variability in SM and overestimated the high-frequency variability for the 50–100 cm depth. Based on these evaluation results, the three ESM-independent products were finally recommended for future applications because of their better performances than the ESM-dependent ones. Despite uncertainties in the raw SM datasets and fusion methods, these hybrid products create added value over existing SM datasets because of the performance improvement and harmonized spatial, temporal, and vertical coverages, and they provide a new foundation for scientific investigation and resource management.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2020

Designing accurate emulators for scientific processes using calibration-driven deep models

Abstract Predictive models that accurately emulate complex scientific processes can achieve speed-ups over numerical simulators or experiments and at the same time provide surrogates for improving the subsequent analysis. Consequently, there is a recent surge in utilizing modern machine learning methods to build data-driven emulators. In this work, we study an often overlooked, yet important, problem of choosing loss functions while designing such emulators. Popular choices such as the mean squared error or the mean absolute error are based on a symmetric noise assumption and can be unsuitable for heterogeneous data or asymmetric noise distributions. We propose Learn-by-Calibrating, a novel deep learning approach based on interval calibration for designing emulators that can effectively recover the inherent noise structure without any explicit priors. Using a large suite of use-cases, we demonstrate the efficacy of our approach in providing high-quality emulators, when compared to widely-adopted loss function choices, even in small-data regimes.

97 MATHEMATICS AND COMPUTING↗