Search NASA⌕ Search

SEARCH · Search NASA

Results for “encoding”

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 379 records · Page 21

Advanced Polymer Characterization: Modular Operations for Spectral Alignment by Iterative Compression (MOSAIC)

Matrix-assisted laser desorption/ionization (MALDI) mass spectrometry encodes structural information across diverse homo- and copolymer ensembles, yet decrypting these spectra requires a systematic analytical approach. We introduce Modular Operations for Spectral Alignment by Iterative Compression (MOSAIC)─a general cipher algorithm that applies modular arithmetic to filter monomer-derived mass contributions and cluster MALDI peaks by nonconstitutional repeating units (non-CRUs). MOSAIC performs sequential modular operations using monomer mass differences as base units to compress complex spectral data, revealing end-group distributions and comonomer incorporation. As a demonstration, we applied MOSAIC to five copolymers formed by two different polymerization mechanisms. Furthermore, the resulting remainder–mass plots clearly resolve polymer homologs with distinct non-CRUs into visually apparent clusters, enabling intuitive assignment of mass spectral features.

Wang, Hanlin M. [University of Illinois at Urbana−↗

Water, Solute, and Ion Transport in De Novo-Designed Membrane Protein Channels

Biological organisms engineer peptide sequences to fold into membrane pore proteins capable of performing a wide variety of transport functions. Synthetic de novo-designed membrane pores can mimic this approach to achieve a potentially even larger set of functions. Here, in this work, we explore water, solute, and ion transport in three de novo designed β-barrel membrane channels in the 5–10 Å pore size range. We show that these proteins form passive membrane pores with high water transport efficiencies and size rejection characteristics consistent with the pore size encoded in the protein structure. Ion conductance and ion selectivity measurements also show trends consistent with the pore size, with the two larger pores showing weak cation selectivity. MD simulations of water and ion transport and solute size exclusion are consistent with the experimental trends and provide further insights into structure–function correlations in these membrane pores.

59 BASIC BIOLOGICAL SCIENCES↗

Rethinking Suicide Thi4 Thiazole Synthases: Comparative Genomic Insights and Pilot Functional Evidence

Suicide thiazole synthases (Thi4) are mononuclear metal enzymes that form the thiazole moiety of thiamin from NAD + , glycine, and a sulfur atom that is stripped from an active-site cysteine residue, causing enzyme inactivation. Comparative genomic analysis shows that prokaryotic Thi4 genes often cluster on the chromosomal regions encoding ThiS, ThiF, and other proteins that can produce, relay, or use persulfide or thiocarboxylate sulfur. These recurring genomic associations raise the possibility that, in some microorganisms, suicide Thi4s may interact with sulfur-relay systems, i.e., they can possibly operate in a nonsuicide mode. This proof-of-concept study explores this possibility via complementation assays using Escherichia coli as a heterologous platform. A representative bacterial Thi4 that clustered with thiS and thiF complemented an E. coli ΔthiG (thiazole auxotroph) single mutant better than a ΔthiG ΔthiF ΔthiS triple mutant. Although (in)direct sulfur transfer could not be assessed in the scope of our investigation, the initial results suggest a dependence on host sulfur relay components, consistent with predicted interactions with the host sulfide transfer chain. Collectively, this new perspective provides a useful guide for future biochemical studies on alternative modes of action for “suicide Thi4s” and accessory proteins.

Bacteria↗

A Chimeric LBT-GFP Biosensor Exhibits Antithetical Fluorescence Responses to Ca 2+ and Dy 3+ Binding

Rare earth elements (REEs) are critical components in emerging technologies, but their mining and refining processes are often laborious, costly, and environmentally damaging. Developing green and efficient separation methods for REEs is crucial. Biomolecular approaches using lanthanide-binding proteins and peptides show promise for selective REE extraction and separation. In this study, we present the design and characterization of a genetically encoded fluorescence indicator (GEFI) construct that combines a superfolder green fluorescent protein (sfGFP) with a dual lanthanide-binding tag (2×dLBT). The 2×dLBT insert induces conformational changes in sfGFP upon lanthanide binding, modulating the fluorescence intensity. The sfGFP-2×dLBT biosensor exhibited distinct fluorescence responses to different lanthanide ions, with the highest dynamic range observed for heavy REEs like dysprosium (Dy 3+ ). Interestingly, the sensor displayed an antithetical response, where low concentrations of lanthanides initially quenched the fluorescence, but higher concentrations led to a significant fluorescence increase (1.5-fold). The Ca 2+ ion on the other hand showed only a dose-dependent quenching of the fluorescence response. Based on these observations, the biphasic response of the biosensor to lanthanides was eliminated by pretreating the sensor with calcium, which further expanded the dynamic range up to 3-fold for Dy 3+ . The lanthanide-selective and concentration-dependent fluorescence changes of the sfGFP-2×dLBT biosensor demonstrate its potential as a platform for developing specific sensors for various REEs. These sensors could enable rapid and cost-effective determination of REE composition in complex mixtures, facilitating the separation and recovery of critical REEs from electronic waste and other REE-containing sources.

59 BASIC BIOLOGICAL SCIENCES↗

Quorum-driven microbial consortium for Bioplastic production from agro-waste

Microbial consortia have high relevance in natural environments. Here we present the production of polyhydroxyalkanoates (PHA) from agro-industrial residues by a synthetic interkingdom consortium formed by the saprotrophic fungus Ophiostoma piceae CECT 20146, which encodes a wide range of lignocellulolytic enzymes, and a natural PHA producer, Pseudomonas putida KT2440. Two agro-industrial residues were utilized: Brewer's Spent Grain (BSG) as a carbon/nitrogen source and biofilm scaffold and waste cooking oil (WCO) as a carbon source for PHA synthesis. Through biochemistry, microscopy, and omics analyses, it is shown that P. putida accumulates up to 40.2% of intracellular PHA when the quorum sensing molecule, farnesol (naturally produced by O. piceae) is added, thanks to the increased proliferation of P. putida cells. An interactive Shiny application has also been developed for an easy visualization and comprehension of all the transcriptomics and metabolomics data: https://jgf-bioinformatics.shinyapps.io/Visualization_app/. These results support the increased PHA production of the consortium by an induction of gene phaG, which redirects intermediaries of the fatty acid biosynthesis to PHA precursors, and the repression of the PHA depolymerase phaZ in P. putida. The trophic interaction between microorganisms seems to rely on the citric acid produced by O. piceae and the glycerol liberated from WCO, which can both be consumed by P. putida. Bioreactor scale-up experiments allowed a 3.3-fold increase in the PHA concentration in the consortium (6.7 g·L-1) without pretreatment or sterilization of the substrates, laying the groundwork for the implementation of an industrial consolidated bioprocess (CBP).

Bacteria↗

Engineered Accumulation of Protocatechuate in Corn Biomass to Enhance Biomanufacturing

The in-planta accumulation of coproducts in crops can enhance the value of lignocellulosic biomass and facilitate a sustainable bioeconomy. Corn stover represents a major renewable source of lignocellulose for the production of advanced biofuels and bioproducts. In this study, we engineered corn with a bacterial gene encoding a dehydroshikimate dehydratase (QsuB) to overproduce protocatechuate (DHBA). Transgenic corn lines accumulate up to 2.9% DHBA on a dry weight basis in leaf and stem biomass. DHBA occurs in the form of glucosides that are extractable from biomass using aqueous methanol as the solvent. The analysis of lignin did not show any evidence for the incorporation of DHBA; however, an increase in the lignin syringyl to guaiacyl ratio and a higher relative abundance of p-coumarate groups compared with total lignin units were observed in QsuB-modified corn. Alkaline hydrolysates prepared from QsuB corn were enriched in DHBA compared to the hydrolysates obtained from wild-type biomass, which contained mostly p-coumarate and ferulate. Using engineered Novosphingobium aromaticivorans as a production host, a 375% improvement in 2-pyrone-4,6-dicarboxylate titers was achieved through biological upgrading of alkaline hydrolysates derived from QsuB corn compared to unmodified biomass. Our data demonstrate an engineering strategy to overproduce DHBA in corn that can facilitate sustainable manufacturing of other valuable bioproducts using stover as a feedstock.

2-pyrone-4,6-dicarboxylate↗

Exploring the Structural, Biochemical, and Functional Diversity of Glycoside Hydrolase Family 12 from Penicillium subrubescens

Glycoside hydrolases (GHs) play an essential role in plant biomass degradation and modification for the sustainable production of biochemicals. The filamentous Ascomycete fungus Penicillium subrubescens contains a higher number of GH12 candidates compared to related species. Therefore, we aimed to compare P. subrubescens GH12s for their ability and substrate specificity for plant cell wall polysaccharide degradation and species’ potential as a source of novel enzymes for plant biomass valorization. Our re-evaluated phylogenetic analysis of fungal GH12 members showed that the P. subrubescens GH12s were located in different (new) clades. Biochemical characterization marked PsEglA as an endoglucanase and four other P. subrubescens GH12s (i.e., PsXegA–D) as xyloglucanases. Interestingly, structural features of PsXegD and PsXegE were more comparable to those of Basidiomycete GH12 xyloglucanases with a unique open substrate-binding cleft. PsUegA displayed dual xyloglucanase and endoglucanase activity and also showed distinct structural features. Comparative transcriptome analysis supported the functional diversity of P. subrubescens GH12s in plant biomass degradation. The gene encoding PsUegA was expressed under diverse conditions, suggesting a scouting role for this enzyme.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Engineering Clostridium Tyrobutyricum for High Butanol Production through Induction Expression of Exogenous NADPH-Dependent HBD

Clostridium tyrobutyricum Δ cat 1:: adh E2 is a promising cell factory for butanol production because of its robustness, high butanol tolerance, and minimal butyrate production. However, excessive acetate and ethanol production remains a major bottleneck limiting its butanol yield. Coexpressing an exogenous hbd ( Ck ) encoding the NADPH-dependent 3-hydroxybutyryl-CoA dehydrogenase (HBD) from Clostridium kluyveri with adh E2 could increase the C4 carbon flux, resulting in increased butanol and decreased acetate and ethanol production. However, constitutively overexpressing hbd ( Ck ) in Δ cat 1:: adh E2 shows little improvement in butanol yield, productivity, and selectivity, which might be caused by redox imbalance and growth inhibition. To alleviate this problem, C. tyrobutyricum MΔ cat 1:: adh E2- Pbgal-hbd ( Ck ) with a dynamic expression of hbd ( Ck ) controlled by an inducible promoter was developed. In serum bottle fermentation at 37 °C, when the hbd ( Ck ) expression was induced at 12 h or in the early exponential phase, butanol production increased ∼20% in yield (from 0.22 to 0.27 g/g glucose), 87.5% in productivity (from 0.16 to 0.30 g/L·h), and 52% in selectivity (from 0.46 to 0.70 g/g total products) compared to the control strain without expressing any hbd(Ck), whereas hbd ( Ck ) expression induced at 0 or 24 h in MΔ cat 1:: adh E2-P bgal - hbd ( Ck ) or constitutively in MΔ cat 1:: adh E2-P cat 1- hbd ( Ck ) showed significantly lower butanol yield and productivity. At 25 °C, MΔ cat 1:: adh E2-P bgal - hbd ( Ck ) with 12 h induction produced the highest butanol titer of 23 g/L with 0.32 g/g yield, 0.16 g/L·h productivity, and 0.83 g/g product selectivity due to much reduced acetate formation. Subsequent scale-up to a stirred-tank bioreactor at 37 °C increased productivity to 0.39 g/L·h while also achieving high butanol titer (21.8 g/L), yield (0.30 g/g), and selectivity (0.67 g/g). The optimized induction timing resulted in a balanced NAD(P)H pool, effectively channeling substrates toward butanol biosynthesis. It was concluded that the timing for hbd ( Ck ) expression was critical as it affected glucose catabolism, cell growth, redox balance, and carbon flux distribution. These findings underscore the potential of dynamic metabolic regulation to overcome bottlenecks in biobutanol production, providing a scalable and economically viable bioprocess for industrial application.

Clostridium tyrobutyricum↗

Quantitative Dissection of Agrobacterium Virulence to Generate a Synthetic Ti Plasmid

Agrobacterium is not only a costly plant pathogen but is also an essential tool for plant transformation. Though Agrobacterium-mediated transformation (AMT) has been heavily studied, its polygenic nature and complex transcriptional regulation make identification of the genetic basis of transformational efficiency difficult through traditional genetic and bioinformatic approaches. Here, we use a bottom-up synthetic approach to systematically engineer the tumor-inducing plasmid (pTi), wherein the majority of virulence machinery is encoded. Using a validated toolkit to control Agrobacterium gene expression in planta, we perform a quantitative dissection of AMT to investigate the contributions of critical vir-genes at different expression levels. We construct a synthetic pTi capable of transient plant and stable fungal transformation and characterize bottlenecks and solutions for complex polygenic synthetic pTi designs. Our reductionist approach demonstrates how bottom-up engineering can be used to dissect and elucidate the genetic underpinnings of complex biological traits, laying the foundation for future engineering to establish full synthetic control over the critical process of AMT.

Agrobacterium-mediated transformation↗

Selective Depolymerization for Sculpting Polymethacrylate Molecular Weight Distributions

Chain-end reactivation of polymethacrylates generated by reversible-deactivation radical polymerization (RDRP) has emerged as a powerful tool for triggering depolymerization at significantly milder temperatures than those traditionally employed. In this study, we demonstrate how the facile depolymerization of poly(butyl methacrylate) (PBMA) can be leveraged to selectively skew the molecular weight distribution (MWD) and predictably alter the viscoelastic properties of blended PBMA mixtures. By mixing polymers with thermally active chain ends with polymers of different molecular weights and inactive chain ends, the MWD of the blends can be skewed to be high or low by selective depolymerization. This approach leads to the counterintuitive principle of the “destructive strengthening” of a material. As a result, we demonstrate, as a proof of concept, the encryption of information within polymer mixtures by linking Morse code with the MWDs before and after selective depolymerization, allowing for the encoding of data within blends of synthetic macromolecules.

36 MATERIALS SCIENCE↗

Covalent Drug Binding in Live Cells Monitored by Mid-Infrared Quantum Cascade Laser Spectroscopy: Photoactive Yellow Protein as a Model System

The detection of drug-target interactions in live cells enables analysis of therapeutic compounds in a native cellular environment. Recent advances in spectroscopy and molecular biology have facilitated the development of genetically encoded vibrational probes like nitriles that can sensitively report on molecular interactions. Nitriles are powerful tools for measuring electrostatic environments within condensed media like proteins, but such measurements in live cells have been hindered by low signal-to-noise ratios. In this study, we design a spectrometer based on a double-beam quantum cascade laser (QCL)-based transmission infrared (IR) source with balanced detection that can significantly enhance sensitivity to nitrile vibrational probes embedded in proteins within cells compared to a conventional FTIR spectrometer. Here, using this approach, we detect small-molecule binding in Escherichia coli, with particular focus on the interaction between para-Coumaric acid (pCA) and nitrile-incorporated photoactive yellow protein (PYP). This system effectively serves as a model for investigating covalent drug binding in a cellular environment. Notably, we observe large spectral shifts of up to 15 cm –1 for nitriles embedded in PYP between the unbound and drug-bound states directly within bacteria, in agreement with observations for purified proteins. Such large spectral shifts are ascribed to the changes in the hydrogen-bonding environment around the local environment of nitriles, accurately modeled through high-level molecular dynamics simulations using the AMOEBA force field. Our findings underscore the QCL spectrometer’s ability to enhance sensitivity for monitoring drug–protein interactions, offering new opportunities for advanced methodologies in drug development and biochemical research.

chromophores↗

Circularity in Sequence-Controlled Copolyamides Enabled by Regioselective Enzymatic Hydrolysis

Sequence-controlled polymers enable precise control over macromolecular structures and function, but both their synthesis and end-of-life management remain fundamental challenges. Achieving high sequence fidelity is synthetically demanding, and conventional depolymerization methods lack regioselectivity, leading to irreversible loss of encoded molecular information and limiting polymer circularity. Enzymatic catalysis offers a potential solution by combining substrate specificity with selective bond cleavage. Here, we report the synthesis, characterization, and regioselective enzymatic depolymerization of poly- (X,AMA), a sequence-controlled copolyamide composed of alternating hexamethylenediamine−adipic acid (MA) and pxylylenediamine− adipic acid (XA) repeat units. Poly(X,AMA) was synthesized via solid-state polycondensation (SSP) of sequence-defined oligomers, enabling precise control over repeat-unit order. Polymer microstructure and sequence fidelity were confirmed by 13 C NMR spectroscopy and MALDI−TOF mass spectrometry. Comparison with a statistical copolymer analogue and Nylon-66 demonstrated pronounced differences in crystallinity, morphology, and thermal behavior arising from sequence control. Screening of 96 Nylon hydrolase homologues against poly(X,AMA) revealed strongly enzyme-dependent depolymerization profiles. While tetrad formation was generally favored, enzymes displayed pronounced sequence selectivity, preferentially releasing distinct sequence-defined tetrads XAMA or MAXA. SSP of sequence-defined tetrad MAXA produced a copolyamide with near identical monomer ordering as poly(X,AMA). Computational modeling of enzyme−substrate complexes identified structural features consistent with the observed regioselectivity. Together, these results establish selective enzymatic depolymerization as a viable strategy for the circular recycling of sequence-controlled polymers and provide a foundation for the rational engineering of enzymes for programmable polymer deconstruction.

Amides↗

Unsteady Land-Sea Breeze Circulations in the Presence of a Synoptic Pressure Forcing

Unsteady land-sea breezes (LSBs) that result from time-varying surface temperature contrasts Δθ(t) are explored in the presence of a constant synoptic pressure forcing, M g , oriented from sea to land (α = 0°) or land to sea (α = 180°). Large eddy simulations reveal the development of four distinctive regimes, depending on the joint interaction between M g , α, and Δθ(t) in modulating the fine-scale dynamics. Time lags, computed as the shifts that maximize correlation coefficients of the velocity between the unsteady and the corresponding steady scenarios at Δθ = Δθ max , are found to be significant and to extend 2 hr longer for α = 0° compared to α = 180°. These diurnal dynamics result in nonequilibrium conditions that are significantly affected by the flow history, and that behave differently over the two patches for the different α’s. Turbulence is found to be out of equilibrium with the mean flow, and the mean itself is found to be out of equilibrium with the thermal forcing. The sea surface heat flux is consistently more sensitive than its land counterpart to the time-varying external forcing Δθ(t), and more so for synoptic forcing from land to sea (α = 180°). Hence, although the land reaches equilibrium faster, the sea patch is found to exert a stronger control on the turbulence-mean flow equilibrium response. Finally, the vertical velocity profile at the shore and shore-normal velocity transects at the first grid level are shown to encode the multiscale regimes of the LSBs evolution and can thus be used to identify these regimes using k-means clustering.

58 GEOSCIENCES↗

Spatio–Temporal Machine Learning for Regional to Continental Scale Terrestrial Hydrology

Integrated hydrologic models can simulate coupled surface and subsurface processes but are computationally expensive to run at high resolutions over large domains. Here we develop a novel deep learning model to emulate subsurface flows simulated by the integrated ParFlow–CLM model across the contiguous US. We compare convolutional neural networks like ResNet and UNet run autoregressively against our novel architecture called the Forced SpatioTemporal RNN (FSTR). The FSTR model incorporates separate encoding of initial conditions, static parameters, and meteorological forcings, which are fused in a recurrent loop to produce spatiotemporal predictions of groundwater. We evaluate the model architectures on their ability to reproduce 4D pressure heads, water table depths, and surface soil moisture over the contiguous US at 1 km resolution and daily time steps over the course of a full water year. The FSTR model shows superior performance to the baseline models, producing stable simulations that capture both seasonal and event–scale dynamics across a wide array of hydroclimatic regimes. The emulators provide over 1,000× speedup compared to the original physical model, which will enable new capabilities like uncertainty quantification and data assimilation for integrated hydrologic modeling that were not previously possible. Our results demonstrate the promise of using specialized deep learning architectures like FSTR for emulating complex process–based models without sacrificing fidelity.

54 ENVIRONMENTAL SCIENCES↗

Data Imbalance, Uncertainty Quantification, and Transfer Learning in Data‐Driven Parameterizations: Lessons From the Emulation of Gravity Wave Momentum Transport in WACCM

Abstract Neural networks (NNs) are increasingly used for data‐driven subgrid‐scale parameterizations in weather and climate models. While NNs are powerful tools for learning complex non‐linear relationships from data, there are several challenges in using them for parameterizations. Three of these challenges are (a) data imbalance related to learning rare, often large‐amplitude, samples; (b) uncertainty quantification (UQ) of the predictions to provide an accuracy indicator; and (c) generalization to other climates, for example, those with different radiative forcings. Here, we examine the performance of methods for addressing these challenges using NN‐based emulators of the Whole Atmosphere Community Climate Model (WACCM) physics‐based gravity wave (GW) parameterizations as a test case. WACCM has complex, state‐of‐the‐art parameterizations for orography‐, convection‐, and front‐driven GWs. Convection‐ and orography‐driven GWs have significant data imbalance due to the absence of convection or orography in most grid points. We address data imbalance using resampling and/or weighted loss functions, enabling the successful emulation of parameterizations for all three sources. We demonstrate that three UQ methods (Bayesian NNs, variational auto‐encoders, and dropouts) provide ensemble spreads that correspond to accuracy during testing, offering criteria for identifying when an NN gives inaccurate predictions. Finally, we show that the accuracy of these NNs decreases for a warmer climate (4 × CO 2 ). However, their performance is significantly improved by applying transfer learning, for example, re‐training only one layer using ∼1% new data from the warmer climate. The findings of this study offer insights for developing reliable and generalizable data‐driven parameterizations for various processes, including (but not limited to) GWs.

54 ENVIRONMENTAL SCIENCES↗

Harnessing the Power of Machine Learning and Omics to Identify Environmental Regulation on Microbial Functional Composition for Soil C, N, and P Cycling

Microbial enzyme-mediated soil organic matter (SOM) decomposition regulates many key ecosystem functions, such as elemental cycling, soil carbon sequestration, and soil fertility. However, representing microbial processes in Earth system models (ESMs) remains challenging due to a limited understanding of the spatial patterns of diverse microbial functions responsible for soil carbon (C), nitrogen (N), and phosphorus (P) cycling as well as the underlying mechanisms regulating their relative abundances across various environments. We collected published metagenomics data across the continental US (CONUS) to identify hundreds of microbial genes involved in soil C, N, and P cycling and grouped them into eight enzyme functional classes (EFCs). Each EFC represented a group of gene-encoded potential enzymes that decompose similar soil compounds. By integrating the abundances of omics-informed EFCs with the corresponding environmental information, we trained a machine learning (ML) model to identify key edaphic, climate, and vegetation factors regulating the abundances of each EFC. Quantitative analysis of effects of these factors revealed that the spatial distribution of eight EFCs for soil C, N, and P cycling across CONUS reflected potential resource optimization strategies of microbial communities under nutrient limitation, preferential organic-mineral associations, and climatological stresses. This insight, together with the interpreted ML tool and the CONUS-level benchmark for EFCs abundances, paves the way for parameterizing environmental-regulated microbial functional dynamics in biogeochemical models.

machine learning↗

Simulating Atmospheric Processes in Earth System Models and Quantifying Uncertainties With Deep Learning Multi‐Member and Stochastic Parameterizations

Abstract Deep learning is a powerful tool to represent subgrid processes in climate models, but many application cases have so far used idealized settings and deterministic approaches. Here, we develop stochastic parameterizations with calibrated uncertainty quantification to learn subgrid convective and turbulent processes and surface radiative fluxes of a superparameterization embedded in an Earth System Model (ESM). We explore three methods to construct stochastic parameterizations: (a) a single Deep Neural Network (DNN) with Monte Carlo Dropout; (b) a multi‐member parameterization; and (c) a Variational Encoder Decoder with latent space perturbation. We show that the multi‐member parameterization improves the representation of convective processes, especially in the planetary boundary layer, compared to individual DNNs. The respective uncertainty quantification illustrates that methods (b) and (c) are advantageous compared to a dropout‐based DNN parameterization regarding the spread of convective processes. Hybrid simulations with our best‐performing multi‐member parameterizations remained challenging and crash within the first days. Therefore, we develop a pragmatic partial coupling strategy relying on the superparameterization for condensate emulation. Partial coupling reduces the computational efficiency of hybrid Earth‐like simulations but enables model stability over 5 months with our multi‐member parameterizations. However, our hybrid simulations exhibit biases in thermodynamic fields and differences in precipitation patterns. Despite this, the multi‐member parameterizations enable improvements in reproducing tropical extreme precipitation compared to a traditional convection parameterization. Despite these challenges, our results indicate the potential of a new generation of multi‐member machine learning parameterizations leveraging uncertainty quantification to improve the representation of stochasticity of subgrid effects.

Behrens, Gunnar [Deutsches Zentrum für Luft‐ und R↗

Learning Constitutive Relations From Soil Moisture Data via Physically Constrained Neural Networks

Abstract The constitutive relations of the Richardson‐Richards equation encode the macroscopic properties of soil water retention and conductivity. These soil hydraulic functions are commonly represented by models with a handful of parameters. The limited degrees of freedom of such soil hydraulic models constrain our ability to extract soil hydraulic properties from soil moisture data via inverse modeling. We present a new free‐form approach to learning the constitutive relations using physically constrained neural networks. We implemented the inverse modeling framework in a differentiable modeling framework, JAX, to ensure scalability and extensibility. For efficient gradient computations, we implemented implicit differentiation through a nonlinear solver for the Richardson‐Richards equation. We tested the framework against synthetic noisy data and demonstrated its robustness against varying magnitudes of noise and degrees of freedom of the neural networks. We applied the framework to soil moisture data from an upward infiltration experiment and demonstrated that the neural network‐based approach was better fitted to the experimental data than a parametric model and that the framework can learn the constitutive relations.

54 ENVIRONMENTAL SCIENCES↗