Representation Of Heterogeneity Research Articles

Cervical extracellular matrix hydrogel optimizes tumor heterogeneity of cervical squamous cell carcinoma organoids.

Cervical cancer, primarily squamous cell carcinoma, is the most prevalent gynecologic malignancy. Organoids can mimic tumor development in vitro, but current Matrigel inaccurately replicates the tissue-specific microenvironment. This limitation compromises the accurate representation of tumor heterogeneity. We collected para-cancerous cervical tissues from patients diagnosed with cervical squamous cell carcinoma (CSCC) and prepared uterine cervix extracellular matrix (UCEM) hydrogels. Proteomic analysis of UCEM identified several tissue-specific signaling pathways including human papillomavirus, phosphatidylinositol 3-kinase-AKT, and extracellular matrix receptor. Secreted proteins like FLNA, MYH9, HSPA8, and EEF1A1 were present, indicating UCEM successfully maintained cervical proteins. UCEM provided a tailored microenvironment for CSCC organoids, enabling formation and growth while preserving tumorigenic potential. RNA sequencing showed UCEM-organoids exhibited greater similarity to native CSCC and reflected tumor heterogeneity by exhibiting CSCC-associated signaling pathways including virus protein-cytokine, nuclear factor κB, tumor necrosis factor, and oncogenes EGR1, FPR1, and IFI6. Moreover, UCEM-organoids developed chemotherapy resistance. Our research provides insights into advanced organoid technology through native matrix hydrogels.

Polymer Coatings Affect Transport and Remobilization of Colloidal Activated Carbon in Saturated Sand Columns: Implications for In Situ Groundwater Remediation.

Colloidal activated carbon (CAC) is an emerging technology for the in situ remediation of groundwater impacted by per- and polyfluoroalkyl substances (PFAS). In assessing the long-term effectiveness of a CAC barrier, it is crucial to evaluate the potential of emplaced CAC particles to be remobilized and migrate away from the sorptive barrier. We examine the effect of two polymer stabilizers, carboxymethyl cellulose (CMC) and polydiallyldimethylammonium chloride (PolyDM), on CAC deposition and remobilization in saturated sand columns. CMC-modified CAC showed high mobility in a wide ionic strength (IS) range from 0.1 to 100 mM, which is favorable for CAC delivery at a sufficient scale. Interestingly, the mobility of PolyDM-modified CAC was high at low IS (0.1 mM) but greatly reduced at high IS (100 mM). Notably, significant remobilization (release) of deposited CMC-CAC particles occurred upon the introduction of solution with low IS following deposition at high IS. In contrast, PolyDM-CAC did not undergo any remobilization following deposition due to its favorable interactions with the quartz sand. We further elucidated the CAC deposition and remobilization behaviors by analyzing colloid-collector interactions through the application of Derjaguin-Landau-Verwey-Overbeek theory, and the inclusion of a discrete representation of charge heterogeneity on the quartz sand surface. The classical colloid filtration theory was also employed to estimate the travel distance of CAC in saturated columns. Our results underscore the roles of polymer coatings and solution chemistry in CAC transport, providing valuable guidelines for the design of in situ CAC remediation with maximized delivery efficiency and barrier longevity.

Digital Microscopic Multiphase Heterogeneity Representation and Its Effects on Micromechanics and Cracking Behaviors of Geomaterials

Digital Microscopic Multiphase Heterogeneity Representation and Its Effects on Micromechanics and Cracking Behaviors of Geomaterials

IMWMOTE: A novel oversampling technique for fault diagnosis in heterogeneous imbalanced data

In the realm of smart manufacturing, the urgent necessity of surmounting the technical hurdles associated with developing data-driven fault monitoring and intelligent diagnostic technologies for high-end equipment is imminent. In practical engineering applications, the challenging operating conditions and unfavorable environments for data collection often lead to the acquisition of health data that is not only constrained in terms of sample size but also exhibits imbalances and heterogeneity, collectively referred to as Small-Sample, Imbalanced, and Mixed-Class Scenarios. This poses a considerable challenge to the interpretability modeling of faults in high-end equipment, as identified by the scientific community. Prior investigations have highlighted the distinct benefits of resampling techniques in addressing the complexities of imbalanced fault modeling. Nevertheless, the suboptimal performance of current techniques remains a concern, primarily due to ambiguous representations of sample feature heterogeneity and the lack of sophisticated sampling balance strategies. For instance, the absence of tailored strategies and concomitant optimization approaches specifically designed for each heterogeneous feature remains a notable gap. In this study, we undertake a comprehensive analysis of the heterogeneity mechanisms inherent in fault feature data, with a particular emphasis on components like turbine rotor faults. Building upon these insights, we introduce a fault diagnosis framework utilizing the novel oversampling technique based-MWMOTE (IMWMOTE), to address these challenges effectively. This comprehensive framework not only ensures interpretability in classification but also seamlessly integrates an optimization mechanism. Notably, the efficacy and innovation of our proposed approach have been thoroughly verified through an extensive evaluation involving 20 fault diagnosis instances across three distinct categories: turbine rotor, bearings, and gears. The experimental outcomes clearly illustrate that IMWMOTE exhibits remarkable advantages in tackling the intricate challenges associated with imbalanced fault modeling, particularly when dealing with small sample heterogeneity. Given its demonstrated proficiency, the proposed approach holds significant potential in guiding interpretable fault modeling efforts in related domains, thus contributing to the advancement of fault diagnosis techniques.

FLEXR GUI: a graphical user interface for multi-conformer modeling of proteins.

Proteins are well known 'shapeshifters' which change conformation to function. In crystallography, multiple conformational states are often present within the crystal and the resulting electron-density map. Yet, explicitly incorporating alternative states into models to disentangle multi-conformer ensembles is challenging. We previously reported the tool FLEXR, which, within a few minutes, automatically separates conformational signal from noise and builds the corresponding, often missing, structural features into a multi-conformer model. To make the method widely accessible for routine multi-conformer building as part of the computational toolkit for macromolecular crystallography, we present a graphical user interface (GUI) for FLEXR, designed as a plugin for Coot 1. The GUI implementation seamlessly connects FLEXR models with the existing suite of validation and modeling tools available in Coot. We envision that FLEXR will aid crystallographers by increasing access to a multi-conformer modeling method that will ultimately lead to a better representation of protein conformational heterogeneity in the Protein Data Bank. In turn, deeper insights into the protein conformational landscape may inform biology or provide new opportunities for ligand design. The code is open source and freely available on GitHub at https://github.com/TheFischerLab/FLEXR-GUI.

Abstract 6897: Defining critical features for successful preclinical use of glioblastoma patient-derived xenografts (PDX) models for therapeutic assessment and clinical trial development

Abstract Mayo Clinic has developed 111 GBM PDXs from newly diagnosed (72 PDX) and recurrent (39 PDX) patients. This large collection of PDXs provides a broad representation of genetic heterogeneity seen in patients. GBM PDXs that perform well in both in vitro and in vivo settings can be extensively utilized in the preclinical setting for downstream analysis. Based on 20 years of experience, we have defined the attributes that best enable preclinical use of our GBM PDX models. Among them are short-term in vitro culturing performance, large scale genomic evaluation data, benchmarked responses to standards of care, and median time to moribund for untreated orthotopic tumors. PDXs with prolonged survival timelines delay study readouts and, in some situations, confound efficacy results due to age-related animal decline. Setting median survival thresholds as part of the upfront experimental design allows for more timely study readouts that are unaffected by these confounding variables. Our preferred cutoff for the Mayo GBM PDX collection is 70 days median survival (n=64). Using this 64 PDX line panel, regimens expected to double survival would produce results within 4-5 months. This timeframe is well within the range of robust health for athymic nude mice. This 64-line collection features both newly diagnosed (nGBM; n=42) and recurrent (rGBM; n=22) tumors. MGMT promoter methylation status was determined by qMS-PCR (25 methylated, 37 unmethylated, 2 indeterminate). Available WES (n=62) and RNAseq (n=42) data, along with our tissue microarray (n=42), are critical resources for biomarker-based evaluations aimed at defining patient populations that are most likely to benefit from a given therapeutic. From these data, the molecular subtypes available include 26 classical, 14 proneural, and 3 mesenchymal. Twenty-seven of these 64 PDX models have also been successfully cultured from fresh as well as cryopreserved cell stocks to further expand their experimental value. Correlative clinical data including diagnosis and treatments are also available (n=61). Current standards of care for GBM include radiation, temozolomide, and bevacizumab. Defining sensitivity to these therapies is an important first step for evaluating the integration of novel therapeutics with these standards. Since the creation of our GBM PDX collection, efficacy of these therapies has been evaluated as follows: RT 26 PDXs, TMZ 20 PDXs, and Avastin 23 PDXs. The Mayo Clinic GBM PDX collection provides an important platform for preclinical evaluation of novel therapeutics. Through careful selection of PDX lines with favorable growth characteristics and known genetic features, researchers can select the most relevant models for their study. This approach will enable more meaningful preclinical work in the hopes of providing more tangible clinical benefits for GBM patients. Citation Format: Danielle M. Burgenske, Ann C. Mladek, Katrina K. Bakken, Zeng Hu, Brett L. Carlson, Paul A. Decker, Jeanette E. Eckel-Passow, Jann N. Sarkaria. Defining critical features for successful preclinical use of glioblastoma patient-derived xenografts (PDX) models for therapeutic assessment and clinical trial development [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 6897.

An Unstructured Mesh Generation Tool for Efficient High‐Resolution Representation of Spatial Heterogeneity in Land Surface Models

AbstractThis study develops an unstructured mesh generation tool for land surface modeling using a multi‐scale hexagon discrete mesh. The tool can automatically determine the required mesh resolution for different regions based on multi‐objective criteria such as elevation, slope, land cover, and land use. The refined unstructured meshes demonstrate significant enhancement in the representation of spatial heterogeneity. The tool is coupled with the Common Land Model (CoLM) to enable land surface simulations using unstructured meshes. Evaluations focused on runoff, river discharge, and inundation indicate improved model performance compared to traditional structured mesh‐based CoLM simulations under the same computational cost constraints. This tool provides new capabilities for more efficiently capturing localized land surface processes and extreme events.

Learning structural heterogeneity from cryo-electron sub-tomograms with tomoDRGN.

Cryo-electron tomography (cryo-ET) enables observation of macromolecular complexes in their native, spatially contextualized cellular environment. Cryo-ET processing software to visualize such complexes at nanometer resolution via iterative alignment and averaging are well developed but rely upon assumptions of structural homogeneity among the complexes of interest. Recently developed tools allow for some assessment of structural diversity but have limited capacity to represent highly heterogeneous structures, including those undergoing continuous conformational changes. Here we extend the highly expressive cryoDRGN (Deep Reconstructing Generative Networks) deep learning architecture, originally created for single-particle cryo-electron microscopy analysis, to cryo-ET. Our new tool, tomoDRGN, learns a continuous low-dimensional representation of structural heterogeneity in cryo-ET datasets while also learning to reconstruct heterogeneous structural ensembles supported by the underlying data. Using simulated and experimental data, we describe and benchmark architectural choices within tomoDRGN that are uniquely necessitated and enabled by cryo-ET. We additionally illustrate tomoDRGN's efficacy in analyzing diverse datasets, using it to reveal high-level organization of human immunodeficiency virus (HIV) capsid complexes assembled in virus-like particles and to resolve extensive structural heterogeneity among ribosomes imaged in situ.

An extended multi-expert concept lattice-based heterogeneous multi-attribute group decision-making approach

In practical multi-attribute group decision-making (MAGDM) problems, it is common to utilize heterogeneous representation forms to express distinct preference information for different experts, primarily due to their diverse backgrounds. To address the problem of reducing information loss from the aggregation of experts' opinions in heterogeneous MAGDM as well as improving the interpretability of the decision-making process, this paper introduces a concept lattice-based heterogeneous MAGDM approach. The heterogeneous multi-expert formal context is first proposed to capture the heterogeneous evaluation information of alternatives provided by different experts. Then, extended multi-expert concept lattices are constructed to aggregate evaluation information of alternatives by different experts. In this case, all concepts are considered to minimize information loss during the aggregation process. Second, to obtain more reasonable decision results, the distance between the concept intents and the heterogeneous positive and negative ideal solutions is considered, and the alternatives are ranked based on this measure. In addition, the MAGDM process for each alternative is visualized using the extended multi-expert concept lattices. This representation aids in identifying key concepts, their interdependencies, and the overall impact on the decision result. Finally, numerical examples and comparative analysis validate the validity and rationality of the proposed approach in heterogeneous MAGDM.

Edge-Bound Change Detection in Multisource Remote Sensing Images

Detecting changes in multisource heterogeneous images is a great challenge for unsupervised change detection methods. Image-translation-based methods, which transform two images to be homogeneous for comparison, have become a mainstream approach. However, most of them primarily rely on information from unchanged regions, resulting in networks that cannot fully capture the connection between two heterogeneous representations. Moreover, the lack of a priori information and sufficient training data makes the training vulnerable to the interference of changed pixels. In this paper, we propose an edge-oriented generative adversarial network (EO-GAN) for change detection that indirectly translates images using edge information, which serves as a core and stable link between heterogeneous representations. The EO-GAN is composed of an edge extraction network and a reconstructive network. During the training process, we ensure that the edges extracted from heterogeneous images are as similar as possible through supplemented data based on superpixel segmentation. Experimental results on both heterogeneous and homogeneous datasets demonstrate the effectiveness of our proposed method.

Modelling the coupled heterogeneities of the lacustrine microbialite-bearing carbonate reservoir of the Yacoraite Formation (Salta, Argentina)

The Yacoraite Formation is a complex lacustrine microbialite-bearing carbonate formation whose heterogeneity is due to different sedimentary facies associations and several microbialite geobodies. The bi-variate plurigaussian geostatistical method is applied to simulate in parallel these two types of heterogeneity in a 3D reservoir-scale geological model. Each variable is simulated by a complete plurigaussian method based on geological interpretation and field observations and quantifications. They are coupled through the occurrence of the microbialites within the sedimentary facies and associated into combined facies giving a one-glance representation of the reservoir heterogeneity, then used to constrain the porosity simulation.

An interpretable knee replacement risk assessment system for osteoarthritis patients

ObjectiveKnee osteoarthritis (OA) is a complex disease with heterogeneous representations. Although it is modifiable to prevention and early treatment, there still lacks a reliable and accurate prognostic tool. Hence, we aim to develop a quantitative and self-administrable knee replacement (KR) risk stratification system for knee osteoarthritis (KOA) patients with clinical features. MethodA total of 14 baseline features were extracted from 9592 cases in the Osteoarthritis Initiative (OAI) cohort. A survival model was constructed using the Random Survival Forests algorithm. The prediction performance was evaluated with the concordance index (C-index) and average receiver operating characteristic curve (AUC). A three-class KR risk stratification system was built to differentiate three distinct KR-free survival groups. Thereafter, Shapley Additive Explanations (SHAP) was introduced for model explanation. ResultsKR incidence was accurately predicted by the model with a C-index of 0.770 (±0.0215) and an average AUC of 0.807 (±0.0181) with 14 clinical features. Three distinct survival groups were observed from the ten-point KR risk stratification system with a four-year KR rate of 0.79%, 5.78%, and 16.2% from the low, medium, and high-risk groups respectively. KR is mainly caused by pain medication use, age, surgery history, diabetes, and a high body mass index, as revealed by SHAP. ConclusionA self-administrable and interpretable KR survival model was developed, underscoring a KR risk scoring system to stratify KOA patients. It will encourage regular self-assessments within the community and facilitate personalised healthcare for both primary and secondary prevention of KOA.

Enhancing predictions of remedial reagent transport via a vertical groundwater circulation well with high-resolution aquifer characterization

The vertical groundwater circulation well (GCW) is a commonly used technique in contaminated sites to remove secondary contaminants from low permeable zones. Early GCW studies often used simple subsurface hydraulic properties, such as anisotropic homogeneous aquifers or low conductivity lens/blocks, to mimic the complex subsurface heterogeneity. Although studies based on simplified representations of aquifer heterogeneity provide straightforward flow and transport information for engineering design of a GCW, they may over- or under-estimate contaminant fate and transport in the field. The objective of this study is to identify key heterogeneity factors that control the capture zone extension and to examine the extent to which the accuracy of estimated heterogeneity spatial distributions influences the prediction of remedial reagent transport. To achieve these objectives, we utilized Monte Carlo simulation to investigate the extension of the circulation zone in heterogeneous aquifers and to identify the key factors that contribute most to the variability of the circulation zone. Three commonly used geostatistical approaches (equivalent homogeneous, kriging, and highly parameterized methods) were employed to estimate the spatial distributions of key factors. The reliabilities of these estimated fields were evaluated through their remedial reagent transport predictability. The key factor analysis revealed that the mean porosity value, the variance of lnK, and the correlation length of lnK profoundly influence the lateral expansion of the capture zone. Neglecting the aquifer hydraulic conductivity heterogeneity underestimates the extension of the circulation zone and the spread of remedial reagent. Additionally, utilizing a highly parameterized approach to estimate the high-resolution K field can accurately reproduce the key remedial reagent distributions. The concentration arrival time and peak concentration are significantly improved compared to those predictions based on the equivalent homogeneous and kriged K fields.

A nonparametric random effects model for the valuation of forest recreation services: An application to forest sites in Tuscany, Italy

AbstractThis study assesses individuals' preferences for the use of forest sites for recreational purposes by means of the logit‐mixed logit (LML) model. The appeal of the LML is that the analyst does not need to assume any specific functional form for the mixing distributions of random preferences. The empirical analysis generates a data‐driven nonparametric representation of individuals' preference heterogeneity. We apply this approach to data collected using an unlabelled discrete choice experiment (DCE), consisting of three recreational options, two of which are in two hypothetical forest sites. Forest destinations are described by means of six attributes: forest type, signposting, hiking time, access to rivers or lakes, wildlife watch hides for visitors and cost of access. The empirical findings reveal that the signpost for each trail is the attribute for which respondents are on average willing to pay the most (6.565€). Further evidence suggests that respondents have strong positive preferences for those forest sites that offer amenities such as wildlife watching hides and access to rivers or lakes. Finally, the histograms derived from the semi‐parametric LML estimation reveal multimodality of random taste amongst respondents for different hypothetical forest sites.

A brain metastasis liquid biopsy: Where are we now?

Brain metastases remain a challenging and feared complication for patients with cancer and research in this area has lagged behind research into metastases to other organs. Due to their location and the risks associated with neurosurgical biopsies, the biology underpinning brain metastases response to treatment and evolution over time remains poorly understood. Liquid biopsies are proposed to overcome many of the limitations present with tissue biopsies, providing a better representation of tumor heterogeneity, facilitating repeated sampling, and providing a noninvasive assessment of tumor biology. Several different liquid biopsy approaches have been investigated including circulating tumor cells, circulating tumor DNA, extracellular vesicles, and tumor-educated platelets; however, these have generally been less effective in assessing brain metastases compared to metastases to other organs requiring improved techniques to investigate these approaches, studies combining different liquid biopsy approaches and/or novel liquid biopsy approaches. Through this review, we highlight the current state of the art and define key unanswered questions related to brain metastases liquid biopsies.

FTMMR: Fusion Transformer for Integrating Multiple Molecular Representations.

Molecular property prediction has gained substantial attention due to its potential for various bio-chemical applications. Numerous attempts have been made to enhance the performance by combining multiple molecular representations (1D, 2D, and 3D). However, most prior works only merged a limited number of representations or tried to embed multiple representations through a single network without using representation-specific networks. Furthermore, the heterogeneous characteristics of each representation made the fusion more challenging. Addressing these challenges, we introduce the Fusion Transformer for Multiple Molecular Representations (FTMMR) framework. Our strategy employs three distinct representation-specific networks and integrates information from each network using a fusion transformer architecture to generate fused representations. Additionally, we use self-supervised learning methods to align heterogeneous representations and to effectively utilize the limited chemical data available. In particular, we adopt a combinatorial loss function to leverage the contrastive loss for all three representations. We evaluate the performance of FTMMR using seven benchmark datasets, demonstrating that our framework outperforms existing fusion and self-supervised methods.

Hierarchical Graph Capsule Networks for Molecular Function Classification with Disentangled Representations.

In biochemistry, graph structures have been widely used for modeling compounds, proteins, functional interactions, etc. A common task that divides these graphs into different categories, known as graph classification, highly relies on the quality of the representations of graphs. With the advance in graph neural networks, message-passing-based methods are adopted to iteratively aggregate neighborhood information for better graph representations. These methods, though powerful, still suffer from some shortcomings. The first challenge is that pooling-based methods in graph neural networks may sometimes ignore the part-whole hierarchies naturally existing in graph structures. These part-whole relationships are usually valuable for many molecular function prediction tasks. The second challenge is that most existing methods do not take the heterogeneity embedded in graph representations into consideration. Disentangling the heterogeneity will increase the performance and interpretability of models. This paper proposes a graph capsule network for graph classification tasks with disentangled feature representations learned automatically by well-designed algorithms. This method is capable of, on the one hand, decomposing heterogeneous representations to more fine-grained elements, whilst on the other hand, capturing part-whole relationships using capsules. Extensive experiments performed on several public-available biochemistry datasets demonstrated the effectiveness of the proposed method, compared with nine state-of-the-art graph learning methods.

Reservoir Characterization of Jeribe and Euphrates Formations in Qaiyarah OilField

Three-dimensional static model is a crucial key for a reservoir heterogeneity representation and an initial evaluation of the hydrocarbon in place, thus for a success field development, an accurate reservoir characterization is a fingerprint for the decision making. The initial goal of this study is to generate a static model for Jeribe and Euphrates formations in Qaiyarah oil field to investigate the distribution of the discrete and continuous proprieties such as, permeability and water saturation, in sequence, recognizing the value of the reserve to identity the ability of developing this reservoir as a final decision for future investment. The sever hetrogeneity and complexity of Qaiyarah oil field were the main challenges confronted through the static model building, in consequence, this paper repesents a fingerprint for the future studies in this area. The conceptual methodology in this research includes the integration of the well log data and final well reports to construct the three-dimensional model through log interpretation, sequential simulators, and properties upscaling. Initial volumetric calculations were performed mainly based on the petrophysical properties distribution of eight wells and oil water contact level. Accordingly, the hydrocarbon volume in place were estimated in the two pay reservoirs under study Jeribe and Euphrates formations. The results showed that Euphrates Formation has dominated by the largest portion of reverse (554*106 Sm3), while Jeribe Formation has lower volume of oil initially in place about 194 *106 Sm3.

Driver Maneuver Interaction Identification with Anomaly-Aware Federated Learning on Heterogeneous Feature Representations

Driver maneuver interaction learning (DMIL) refers to the classification task with the goal of identifying different driver-vehicle maneuver interactions (e.g., left/right turns). Existing conventional studies largely focused on the centralized collection of sensor data from the drivers' smartphones (say, inertial measurement units or IMUs, like accelerometer and gyroscope). Such a centralized mechanism might be precluded by data regulatory constraints. Furthermore, how to enable an adaptive and accurate DMIL framework remains challenging due to (i) complexity in heterogeneous driver maneuver patterns, and (ii) impacts of anomalous driver maneuvers due to, for instance, aggressive driving styles and behaviors. To overcome the above challenges, we propose AF-DMIL, an Anomaly-aware Federated Driver Maneuver Interaction Learning system. We focus on the real-world IMU sensor datasets (e.g., collected by smartphones) for our pilot case study. In particular, we have designed three heterogeneous representations for AF-DMIL regarding spectral, time series, and statistical features that are derived from the IMU sensor readings. We have designed a novel heterogeneous representation attention network (HetRANet) based on spectral channel attention, temporal sequence attention, and statistical feature learning mechanisms, jointly capturing and identifying the complex patterns within driver maneuver behaviors. Furthermore, we have designed a densely-connected convolutional neural network in HetRANet to enable the complex feature extraction and enhance the computational efficiency of HetRANet. In addition, we have designed within AF-DMIL a novel anomaly-aware federated learning approach for decentralized DMIL in response to anomalous maneuver data. To ease extraction of the maneuver patterns and evaluation of their mutual differences, we have designed an embedding projection network that projects the high-dimensional driver maneuver features into low-dimensional space, and further derives the exemplars that represent the driver maneuver patterns for mutual comparison. Then, AF-DMIL further leverages the mutual differences of the exemplars to identify those that exhibit anomalous patterns and deviate from others, and mitigates their impacts upon the federated DMIL. We have conducted extensive driver data analytics and experimental studies on three real-world datasets (one is harvested on our own) to evaluate the prototype of AF-DMIL, demonstrating AF-DMIL's accuracy and effectiveness compared to the state-of-the-art DMIL baselines (on average by more than 13% improvement in terms of DMIL accuracy), as well as fewer communication rounds (on average 29.20% fewer than existing distributed learning mechanisms).

Representation of atmosphere-induced heterogeneity in land–atmosphere interactions in E3SM–MMFv2

Abstract. In the Energy Exascale Earth System Model (E3SM) Multi-scale Modeling Framework (MMF), where parameterizations of convection and turbulence are replaced by a 2-D cloud-resolving model (CRM), there are multiple options to represent land–atmosphere interactions. Here, we propose three different coupling strategies, namely the (1) coupling of a single land surface model to the global grid (MMF), (2) coupling a single land copy directly to the embedded CRM (SFLX2CRM), and (3) coupling a single copy of land model to each column of the CRM grid (MAML). In the MAML (Multi-Atmosphere Multi-Land) framework, a land model is coupled to CRM at the CRM-grid scale by coupling an individual copy of a land model to each CRM grid. Therefore, we can represent intra-CRM heterogeneity in the land–atmosphere interaction processes. There are 5-year global simulations run using these three coupling strategies, and we find some regional differences but overall small changes with respect to whether a land model is coupled to CRM or a global atmosphere. In MAML, the spatial heterogeneity within CRM induces stronger turbulence, which leads to the changes in soil moisture, surface heat fluxes, and precipitation. However, the differences in the MAML from the other two cases are rather weak, suggesting that the impact of using MAML does not justify the increase in cost.