Search NASA⌕ Search

SEARCH · Search NASA

Results for “process modeling”

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 487 records · Page 27

A computational methodology for learning low-complexity surrogate models of process from experiments or simulations. (Paper 679a)

Costly and/or insufficiently robust simulations or experiments can often pose difficulties when their use extends well beyond a single evaluation. This is case with the numerous evaluations of uncertainty quantification, when an algebraic model is needed for optimization, as well as numerous other areas. To overcome these difficulties, we generate an accurate set of algebraic surrogate models of disaggregated process blocks of the experiment or simulation. We developed a method that uses derivative-based and derivative-free optimization alongside machine learning and statistical techniques to generate the set of surrogate models using data sampled from experiments or detailed simulations. Our method begins by building a low-complexity surrogate model for each block from an initial sample set. The model is built using a best subset technique that leverages a mixed-integer linear problem formulation to allow for very large initial basis sets. The models are then tested, exploited, and improved through the use of derivative-free solvers to adaptively sample new simulation or experimental points. The sets of surrogate models from each disaggregated process block are then combined with heat and mass balances around each disaggregated block to generate a full algebraic model of the process. The full model can be used for cheap and accurate evaluations of the original simulation or experiment or combined with design specifications and an objective for nonlinear optimization.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

Bayesian D‐Optimal Designs for Gaussian Process Surrogate Models

Computer experiments often employ space-filling strategies to create surrogate models with strong predictive performance. The impact of model parameter estimation for Gaussian process surrogates, however, is often overlooked. Obtaining a better initial estimate of the covariance lengthscale parameter, θ, can greatly improve the resulting Gaussian process fit through more effective sequential acquisitions during active learning. In this work, we propose a novel initial design maximizing the Bayesian D-optimality criterion of the Gaussian process lengthscale parameter. Previously published results have shown the emphasis on lengthscale estimation to be promising, but relied on an empirically driven design creation process. Our Bayesian D-optimal designs are rooted in information theory and lead to more informative sequential acquisitions by improving lengthscale estimation. In many cases, these gains eventually result in better surrogates than those seeded with space-filling initial designs. Furthermore, Bayesian D-optimal designs can be tailored to either isotropic or anisotropic covariance structures, and the Bayesian framework enables the inclusion of prior knowledge in the design process, offering greater flexibility and adaptability. Through several simulation studies, we demonstrate the advantages of Bayesian D-optimal designs in terms of both lengthscale estimation accuracy and predictive performance during active learning.

Bayesian experimental design↗

Feasibilities of a Coal-Biomass to Liquids Plant in Southern West Virginia

This project has generated comprehensive and realistic results of feasibilities for a coal-biomass to liquids (CBTL) plant in southern West Virginia; and evaluated the sensitivity of the analyses to various anticipated scenarios and parametric uncertainties. Specifically the project has addressed economic feasibility, technical feasibility, market feasibility, and financial feasibility. In the economic feasibility study, a multi-objective siting model was developed and was then used to identify and rank the suitable facility sites. Spatial models were also developed to assess the biomass and coal feedstock availabilities and economics. Environmental impact analysis was conducted mainly to assess life cycle analysis and greenhouse gas emission. Uncertainty and sensitivity analysis were also investigated in this study. Sensitivity analyses on required selling price (RSP) and greenhouse gas (GHG) emissions of CBTL fuels were conducted according to feedstock availability and price, biomass to coal mix ratio, conversion rate, internal rate of return (IRR), capital cost, operational and maintenance cost. The study of siting and capacity showed that feedstock mixed ratio limited the CBTL production. The price of coal had a more dominant effect on RSP than that of biomass. Different mix ratios in the feedstock and conversion rates led to RSP ranging from $104.3 - $157.9/bbl. LCA results indicated that GHG emissions ranged from 80.62 kg CO 2 eq to 101.46 kg CO2 eq/1,000 MJ of liquid fuel at various biomass to coal mix ratios and conversion rates if carbon capture and storage (CCS) was applied. Most of water and fossil energy were consumed in conversion process. Compared to petroleum-derived-liquid fuels, the reduction in GHG emissions could be between -2.7% and 16.2% with CBTL substitution. As for the technical study, three approaches of coal and biomass to liquids, direct, indirect and hybrid, were considered in the analysis. The process models including conceptual design, process modeling and process validation were developed and validated for different cases. Equipment design and capital costs were investigated on capital coast estimation and economical model validation. Material and energy balances and techno-economic analysis on base case were conducted for evaluation of projects. Also, sensitives studies of direct and indirect approaches were both used to evaluate the CBTL plant economic performance. In this study, techno-economic analysis were conducted in Aspen Process Economic Analyzer (APEA) environment for indirect, direct, and hybrid CBTL plants with CCS based on high fidelity process models developed in Aspen Plus and Excel. The process thermal efficiency ranges from 45% to 67%. The break-even oil price ranges from $86.1 to $100.6 per barrel for small scale (10000 bbl/day) CBTL plants and from $65.3 to $80.5 per barrel for large scale (50000 bbl/day) CBTL plants. Increasing biomass/coal ratio from 8/92 to 20/80 would increase the break-even oil price of indirect CBTL plant by $3/bbl and decrease the break-even oil price of direct CBTL plant by about $1/bbl. The order of carbon capture penalty is direct > indirect > hybrid. The order of capital investment is hybrid (with or without shale gas utilization) > direct (without shale gas utilization) > indirect > direct (with shale gas utilization). The order of thermal efficiency is direct > hybrid > indirect. The order of break-even oil price is hybrid (without shale gas utilization) > direct (without shale gas utilization) > hybrid (with shale gas utilization) > indirect > direct (with shale gas utilization).

01 COAL, LIGNITE, AND PEAT↗

Process Cost Modeling for Multi-Disciplinary Design Optimization

For early design concepts, the conventional approach to cost is normally some kind of parametric weight-based cost model. There is now ample evidence that this approach can be misleading and inaccurate. By the nature of its development, a parametric cost model requires historical data and is valid only if the new design is analogous to those for which the model was derived. Advanced aerospace vehicles have no historical production data and are nowhere near the vehicles of the past. Using an existing weight-based cost model would only lead to errors and distortions of the true production cost. This report outlines the development of a process-based cost model in which the physical elements of the vehicle are costed according to a first-order dynamics model. This theoretical cost model, first advocated by early work at MIT, has been expanded to cover the basic structures of an advanced aerospace vehicle. Elemental costs based on the geometry of the design can be summed up to provide an overall estimation of the total production cost for a design configuration. This capability to directly link any design configuration to realistic cost estimation is a key requirement for high payoff MDO problems. Another important consideration in this report is the handling of part or product complexity. Here the concept of cost modulus is introduced to take into account variability due to different materials, sizes, shapes, precision of fabrication, and equipment requirements. The most important implication of the development of the proposed process-based cost model is that different design configurations can now be quickly related to their cost estimates in a seamless calculation process easily implemented on any spreadsheet tool. In successive sections, the report addresses the issues of cost modeling as follows. First, an introduction is presented to provide the background for the research work. Next, a quick review of cost estimation techniques is made with the intention to highlight their inappropriateness for what is really needed at the conceptual phase of the design process. The First-Order Process Velocity Cost Model (FOPV) is discussed at length in the next section. This is followed by an application of the FOPV cost model to a generic wing. For designs that have no precedence as far as acquisition costs are concerned, cost data derived from the FOPV cost model may not be accurate enough because of new requirements for shape complexity, material, equipment and precision/tolerance. The concept of Cost Modulus is introduced at this point to compensate for these new burdens on the basic processes. This is treated in section 5. The cost of a design must be conveniently linked to its CAD representation. The interfacing of CAD models and spreadsheets containing the cost equations is the subject of the next section, section 6. The last section of the report is a summary of the progress made so far, and the anticipated research work to be achieved in the future.

Bao, Han P.↗

A Dirichlet process mixture model for automatic {sup 18}F-FDG PET image segmentation: Validation study on phantoms and on lung and esophageal lesions

Purpose: The aim of this study was to implement a Dirichlet process mixture (DPM) model for automatic tumor edge identification on {sup 18}F-fluorodeoxyglucose positron emission tomography ({sup 18}F-FDG PET) images by optimizing the parameters on which the algorithm depends, to validate it experimentally, and to test its robustness. Methods: The DPM model belongs to the class of the Bayesian nonparametric models and uses the Dirichlet process prior for flexible nonparametric mixture modeling, without any preliminary choice of the number of mixture components. The DPM algorithm implemented in the statistical software package R was used in this work. The contouring accuracy was evaluated on several image data sets: on an IEC phantom (spherical inserts with diameter in the range 10–37 mm) acquired by a Philips Gemini Big Bore PET-CT scanner, using 9 different target-to-background ratios (TBRs) from 2.5 to 70; on a digital phantom simulating spherical/uniform lesions and tumors, irregular in shape and activity; and on 20 clinical cases (10 lung and 10 esophageal cancer patients). The influence of the DPM parameters on contour generation was studied in two steps. In the first one, only the IEC spheres having diameters of 22 and 37 mm and a sphere of the digital phantom (41.6 mm diameter) were studied by varying the main parameters until the diameter of the spheres was obtained within 0.2% of the true value. In the second step, the results obtained for this training set were applied to the entire data set to determine DPM based volumes of all available lesions. These volumes were compared to those obtained by applying already known algorithms (Gaussian mixture model and gradient-based) and to true values, when available. Results: Only one parameter was found able to significantly influence segmentation accuracy (ANOVA test). This parameter was linearly connected to the uptake variance of the tested region of interest (ROI). In the first step of the study, a calibration curve was determined to automatically generate the optimal parameter from the variance of the ROI. This “calibration curve” was then applied to contour the whole data set. The accuracy (mean discrepancy between DPM model-based contours and reference contours) of volume estimation was below (1 ± 7)% on the whole data set (1 SD). The overlap between true and automatically segmented contours, measured by the Dice similarity coefficient, was 0.93 with a SD of 0.03. Conclusions: The proposed DPM model was able to accurately reproduce known volumes of FDG concentration, with high overlap between segmented and true volumes. For all the analyzed inserts of the IEC phantom, the algorithm proved to be robust to variations in radius and in TBR. The main advantage of this algorithm was that no setting of DPM parameters was required in advance, since the proper setting of the only parameter that could significantly influence the segmentation results was automatically related to the uptake variance of the chosen ROI. Furthermore, the algorithm did not need any preliminary choice of the optimum number of classes to describe the ROIs within PET images and no assumption about the shape of the lesion and the uptake heterogeneity of the tracer was required.

60 APPLIED LIFE SCIENCES↗

Incorporating physical constraints into Gaussian process surrogate models (LDRD Project Summary)

This report summarizes work done under the Laboratory Directed Research and Development (LDRD) project titled "Incorporating physical constraints into Gaussian process surrogate models?' In this project, we explored a variety of strategies for constraint implementations. We considered bound constraints, monotonicity and related convexity constraints, Gaussian processes which are constrained to satisfy linear operator constraints which represent physical laws expressed as partial differential equations, and intrinsic boundary condition constraints. We wrote three papers and are currently finishing two others. We developed initial software implementations for some approaches. This report summarizes the work done under this LDRD.

97 MATHEMATICS AND COMPUTING↗

Physics-informed KNN milling stability model with process damping effects

This paper describes a k-nearest neighbors, or KNN, model for milling stability including process damping effects. A physics-based, frequency domain milling stability solution is used to generate the training data, but does not incorporate process damping effects. The data set is then updated using limited tests to capture the process damping behavior. A “stair step” approach is used to select the test points, where a first spindle speed-axial depth combination is selected based on the physics-based stability map, subsequent tests are defined using the previous test result, and data points are updated by knowledge of process damping behavior and the test results. Furthermore, the KNN modeling approach demonstrates the ability to predict both stable and unstable results, including process damping behavior.

42 ENGINEERING↗

Modeling Mesoscale Processes of Scalable Synthesis (Final Report)

The main objective of the center is to develop tools that can help for the understanding of equilibrium and non-equilibrium properties of nanoparticles in aqueous electrolytes. To achieve this goal, we need to develop new computational methodologies that can integrate molecular level description such as the electronic structure analysis to mesoscale description such as generalized hydrodynamics. The challenge is to develop systematic coarse graining techniques that can adaptively choose the level of description needed for the particular task. For example, if bonding breaking is involved, then one should use appropriate models from quantum mechanics to accurately model the time scale and molecular mechanism for the breaking process. However, to model transport process or the process of self-assembly, much less detailed models can be used. There are two main components to this task. One is to develop accurate models that bridge different levels of description. The classical example is the QM/MM model that bridges quantum mechanics and molecular mechanics for studying chemical reactions of large molecules. This work was awarded the Nobel prize this year. For our purpose, we also need to develop methods that bridge molecular and continuum levels of description. Another main task is to develop systematic coarse-graining methods that can give rise to accurate mesoscale and macroscale models, starting from more detailed models at the atomic or molecular scale. None of these ideas are new. By now they have been fashionable topics for quite some time. But neither problem is satisfactorily resolved. The main problem is that uncontrolled approximations are used. As a result, we have very little idea about the validity and accuracy of these models. The main purpose of the center is to exploit the Mori-Zwanzig formalism to achieve the goals stated above. The Mori-Zwanzig formalism is a general tool for eliminating degrees of freedom from existing models. What is attractive about this formalism is its generality. The flip side is that it is also somewhat of a tautology. In order to make it a practical tool, one has to make approximations and this is where the trouble comes. Instead of focusing on Mori-Zwanzig formalism, the PI proposes to study the closely related renormalization group formalism (RG). Like Mori-Zwanzig, RG is a general strategy for eliminating small scale degrees of freedom in a model. Unlike Mori-Zwanzig, RG has the advantage that at least for some problems, one can write a PDE (partial differential equation) type of model for the exact RG flow, without any approximation. This gives us a solid starting point for discussing and analyzing approximations. The PI studied approximate RG flows for turbulent diffusion and related problems. However, extending it to more practical problems proved to be hard. The second line that the PI has pursued is to develop special numerical algorithms for studying particularly important events for dynamics processes in meso-scale systems. The PI succeeded in doing so for the problem of understanding phase transition in metallic systems.

97 MATHEMATICS AND COMPUTING↗

Scalable Statistical Analysis of Gaussian Models for Petascale Spatiotemporal Data (Final Technical Report)

Gaussian process models for spatial and spatiotemporal data are ubiquitous in scientific and engineering applications. Likelihood-based methods for fitting such models, which includes all Bayesian methods, have desirable statistical properties and are thus the methods of choice when feasible. However, calculating the likelihood function even once can be a challenge for N irregularly sited observations from such a process, requiring O(N 2 ) memory and O(N 3 ) complexity. Thus, there is a need for fast and accurate likelihood approximation even for N on the order of 10 5 , not to mention the 10 15 implied by petascale data. The main objective of this research was to develop computational tools for the statistical analysis of massive spatiotemporal datasets based on Gaussian process models. Another important objective was the development of appropriate statistical models for specific spatiotemporal processes arising in the physical sciences.

97 MATHEMATICS AND COMPUTING↗

A process improvement model for software verification and validation

We describe ongoing work at the NASA Independent Verification and Validation (IV&V) Facility to establish a process improvement model for software verification and validation (V&V) organizations. This model, similar to those used by some software development organizations, uses measurement-based techniques to identify problem areas and introduce incremental improvements. We seek to replicate this model for organizations involved in V&V on large-scale software development projects such as EOS and Space Station. At the IV&V Facility, a university research group and V&V contractors are working together to collect metrics across projects in order to determine the effectiveness of V&V and improve its application. Since V&V processes are intimately tied to development processes, this paper also examines the repercussions for development organizations in large-scale efforts.

Callahan, John↗

A process improvement model for software verification and validation

We describe ongoing work at the NASA Independent Verification and Validation (IV&V) Facility to establish a process improvement model for software verification and validation (V&V) organizations. This model, similar to those used by some software development organizations, uses measurement-based techniques to identify problem areas and introduce incremental improvements. We seek to replicate this model for organizations involved in V&V on large-scale software development projects such as EOS and space station. At the IV&V Facility, a university research group and V&V contractors are working together to collect metrics across projects in order to determine the effectiveness of V&V and improve its application. Since V&V processes are intimately tied to development processes, this paper also examines the repercussions for development organizations in large-scale efforts.

Callahan, John↗

In-Situ Calibrated Digital Process Twin Models for Resource Efficient Manufacturing

The chief objective of manufacturing process improvement efforts is to significantly minimize process resources such as time, cost, waste, and consumed energy while improving product quality and process productivity. This paper presents a novel physics-informed optimization approach based on artificial intelligence (AI) to generate digital process twins (DPTs). The utility of the DPT approach is demonstrated in the case of finish machining of aerospace components made from gamma titanium aluminide alloy (γ-TiAl). This particular component has been plagued with persistent quality defects, including surface and sub-surface cracks, which adversely affect resource efficiency. Previous process improvement efforts have been restricted to anecdotal post-mortem investigation and empirical modeling, which fail to address the fundamental issue of how and when cracks occur during cutting. In this work, the integration of in-situ process characterization with modular physics-based models is presented, and machine learning algorithms are used to create a DPT capable of reducing environmental and energy impacts while significantly increasing yield and profitability. Based on the preliminary results presented here, we report an improvement in the overall embodied energy efficiency of over 84%, 93% in process queuing time, 2% in scrap cost, and 93% in queuing cost has been realized for γ-TiAl machining using our novel approach.

42 ENGINEERING↗

Modeling Coupled Processes in Clay Formations for Radioactive Waste Disposal

As a result of the termination of the Yucca Mountain Project, the United States Department of Energy (DOE) has started to explore various alternative avenues for the disposition of used nuclear fuel and nuclear waste. The overall scope of the investigation includes temporary storage, transportation issues, permanent disposal, various nuclear fuel types, processing alternatives, and resulting waste streams. Although geologic disposal is not the only alternative, it is still the leading candidate for permanent disposal. The realm of geologic disposal also offers a range of geologic environments that may be considered, among those clay shale formations. Figure 1-1 presents the distribution of clay/shale formations within the USA. Clay rock/shale has been considered as potential host rock for geological disposal of high-level nuclear waste throughout the world, because of its low permeability, low diffusion coefficient, high retention capacity for radionuclides, and capability to self-seal fractures induced by tunnel excavation. For example, Callovo-Oxfordian argillites at the Bure site, France (Fouche et al., 2004), Toarcian argillites at the Tournemire site, France (Patriarche et al., 2004), Opalinus clay at the Mont Terri site, Switzerland (Meier et al., 2000), and Boom clay at Mol site, Belgium (Barnichon et al., 2005) have all been under intensive scientific investigations (at both field and laboratory scales) for understanding a variety of rock properties and their relations with flow and transport processes associated with geological disposal of nuclear waste. Clay/shale formations may be generally classified as indurated and plastic clays (Tsang et al., 2005). The latter (including Boom clay) is a softer material without high cohesion; its deformation is dominantly plastic. For both clay rocks, coupled thermal, hydrological, mechanical and chemical (THMC) processes are expected to have a significant impact on the long-term safety of a clay repository. For example, the excavation-damaged zone (EDZ) near repository tunnels can modify local permeability (resulting from induced fractures), potentially leading to less confinement capability (Tsang et al., 2005). Because of clay's swelling and shrinkage behavior (depending on whether the clay is in imbibition or drainage processes), fracture properties in the EDZ are quite dynamic and evolve over time as hydromechanical conditions change. To understand and model the coupled processes and their impact on repository performance is critical for the defensible performance assessment of a clay repository. Within the Natural Barrier System (NBS) group of the Used Fuel Disposition (UFD) Campaign at DOE's Office of Nuclear Energy, LBNL's research activities have focused on understanding and modeling such coupled processes. LBNL provided a report in this April on literature survey of studies on coupled processes in clay repositories and identification of technical issues and knowledge gaps (Tsang et al., 2010). This report will document other LBNL research activities within the natural system work package, including the development of constitutive relationships for elastic deformation of clay rock (Section 2), a THM modeling study (Section 3) and a THC modeling study (Section 4). The purpose of the THM and THC modeling studies is to demonstrate the current modeling capabilities in dealing with coupled processes in a potential clay repository. In Section 5, we discuss potential future R&D work based on the identified knowledge gaps. The linkage between these activities and related FEPs is presented in Section 6.

54 ENVIRONMENTAL SCIENCES↗

Coevolution of Machine Learning and Process-Based Modelling to Revolutionize Earth and Environmental Sciences: A Perspective

Machine learning (ML) applications in Earth and environmental sciences (EES) have gained incredible momentum in recent years. However, these ML applications have largely evolved in ‘isolation’ from the mechanistic, process-based modelling (PBM) paradigms, which have historically been the cornerstone of scientific discovery and policy support. In this perspective, we assert that the cultural barriers between the ML and PBM communities limit the potential of ML, and even its ‘hybridization’ with PBM, for EES applications. Fundamental, but often ignored, differences between ML and PBM are discussed as well as their strengths and weaknesses in light of three overarching modelling objectives in EES, (1) nowcasting and prediction, (2) scenario analysis, and (3) diagnostic learning. The paper ponders over a ‘coevolutionary’ approach to model building, shifting away from a borrowing to a co-creation culture, to develop a generation of models that leverage the unique strengths of ML such as scalability to big data and high-dimensional mapping, while remaining faithful to process-based knowledge base and principles of model explainability and interpretability, and therefore, falsifiability.

Saman Razavi↗

Process Model-Based Validation of the Intensification of Biomass Fast Pyrolysis in a Fluidized Bed via Autothermal Operation

A model for the fluidized-bed pyrolysis of biomass is extended to enable the simulation of intensified autothermal operation. In this system, partial oxidation of char and pyrolysate species replaces an external heat source to supply the enthalpy of pyrolysis, greatly increasing the throughput of a given fast pyrolysis reactor. Oxidation reactions are compiled from CRECK and other literature sources and extended to cover the species found in pyrolysis products, including a derivation of the catalytic effect of ash on char combustion from experimental studies at Iowa State University (ISU). Results indicate a roughly 3-fold increase in biomass throughput for a given reactor volume at the pilot scale with minimal loss of valuable products, in agreement with published data from the 3 × 10 -2 m 3 reactor operated by ISU. Additionally, a proposal for a 250 ton/day commercial-scale biomass pyrolysis reactor is analyzed, showing that the same size reactor operated by heating the fluidizing gas externally would have only one-tenth the capacity; if operated by heating the sand externally, the sand would need to circulate at an impractical 0.7 bed volume per minute rate to maintain a 250 ton/day biomass throughput.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Gaussian-process generative model for the QCD equation of state

We develop a generative model for the nuclear matter equation of state at zero net baryon density using the Gaussian process regression method. We impose first-principles theoretical constraints from lattice quantum chromodynamics and hadron resonance gas at high- and low-temperature regions, respectively. By allowing the trained Gaussian process regression model to vary freely near the phase transition region, we generate random smooth crossover equations of state with different speeds of sound that do not rely on specific parametrizations. Here, we explore a collection of experimental observable dependencies on the generated equations of state, which paves the groundwork for future Bayesian inference studies to use experimental measurements from relativistic heavy-ion collisions to constrain the nuclear matter equation of state.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Dynamic Modeling of Process Technologies for Closed-Loop Water Recovery Systems

Detailed chemical process simulations are a useful tool in designing and optimizing complex systems and architectures for human life support. Dynamic and steady-state models of these systems help contrast the interactions of various operating parameters and hardware designs, which become extremely useful in trade-study analyses. NASA s Exploration Life Support technology development project recently made use of such models to compliment a series of tests on different waste water distillation systems. This paper presents dynamic simulations of chemical process for primary processor technologies including: the Cascade Distillation System (CDS), the Vapor Compression Distillation (VCD) system, the Wiped-Film Rotating Disk (WFRD), and post-distillation water polishing processes such as the Volatiles Removal Assembly (VRA) that were developed using the Aspen Custom Modeler and Aspen Plus process simulation tools. The results expand upon previous work for water recovery technology models and emphasize dynamic process modeling and results. The paper discusses system design, modeling details, and model results for each technology and presents some comparisons between the model results and available test data. Following these initial comparisons, some general conclusions and forward work are discussed.

Allada, Rama Kumar↗

Divergent controls of soil organic carbon between observations and process-based models

Abstract The storage and cycling of soil organic carbon (SOC) are governed by multiple co-varying factors, including climate, plant productivity, edaphic properties, and disturbance history. Yet, it remains unclear which of these factors are the dominant predictors of observed SOC stocks, globally and within biomes, and how the role of these predictors varies between observations and process-based models. Here we use global observations and an ensemble of soil biogeochemical models to quantify the emergent importance of key state factors – namely, mean annual temperature, net primary productivity, and soil mineralogy – in explaining biome- to global-scale variation in SOC stocks. We use a machine-learning approach to disentangle the role of covariates and elucidate individual relationships with SOC, without imposing expected relationships a priori . While we observe qualitatively similar relationships between SOC and covariates in observations and models, the magnitude and degree of non-linearity vary substantially among the models and observations. Models appear to overemphasize the importance of temperature and primary productivity (especially in forests and herbaceous biomes, respectively), while observations suggest a greater relative importance of soil minerals. This mismatch is also evident globally. However, we observe agreement between observations and model outputs in select individual biomes – namely, temperate deciduous forests and grasslands, which both show stronger relationships of SOC stocks with temperature and productivity, respectively. This approach highlights biomes with the largest uncertainty and mismatch with observations for targeted model improvements. Understanding the role of dominant SOC controls, and the discrepancies between models and observations, globally and across biomes, is essential for improving and validating process representations in soil and ecosystem models for projections under novel future conditions.

58 GEOSCIENCES↗