Large Bias Research Articles

On parameter bias in earthquake sequence models using data assimilation

Abstract. The feasibility of physics-based forecasting of earthquakes depends on how well models can be calibrated to represent earthquake scenarios given uncertainties in both models and data. We investigate whether data assimilation can estimate current and future fault states, i.e., slip rate and shear stress, in the presence of a bias in the friction parameter. We perform state estimation as well as combined state-parameter estimation using a sequential-importance resampling particle filter in a zero-dimensional (0D) generalization of the Burridge–Knopoff spring–block model with rate-and-state friction. Minor changes in the friction parameter ϵ can lead to different state trajectories and earthquake characteristics. The performance of data assimilation with respect to estimating the fault state in the presence of a parameter bias in ϵ depends on the magnitude of the bias. A small parameter bias in ϵ (+3 %) can be compensated for very well using state estimation (R2 = 0.99), whereas an intermediate bias (−14 %) can only be partly compensated for using state estimation (R2 = 0.47). When increasing particle spread by accounting for model error and an additional resampling step, R2 increases to 0.61. However, when there is a large bias (−43 %) in ϵ, only state-parameter estimation can fully account for the parameter bias (R2 = 0.97). Thus, simultaneous state and parameter estimation effectively separates the error contributions from friction and shear stress to correctly estimate the current and future shear stress and slip rate. This illustrates the potential of data assimilation for the estimation of earthquake sequences and provides insight into its application in other nonlinear processes with uncertain parameters.

Abstract 1024: Improved cfDNA methylation profiling through correction of misrepaired jagged-ends

Abstract Background: cfDNA methylation profiling allows early cancer detection and tissue-of-origin classification. Recent studies show that end repair (ER) process during double strand library preparation (dsLibP) introduces differential methylation signals between original and repaired sequences. We hypothesized that algorithmic correction of cfDNA jagged-ends could allow superior cfDNA methylation profiling performance, and better preserve double-stranded cfDNA molecules. Methods: We introduce JEEPERS (Jagged-End Error Polishing of Enzymatically misRepaired Sequences), a novel approach for correction of such errors in dsLibP methylation data. Specifically, JEEPERS is not only able to detect and quantify ER-induced errors at JEs, but also to correct them in silico. When using double-stranded unique molecular identifiers, JEEPERS leverages support from 2 sources to identify/correct errors at JEs: 1) complementary strands from individual DNA duplexes, and 2) sibling reads from other, non-jagged cfDNA families. JEEPERS relies on the 5’>3’ polarity of ER-associated errors and JE length profiles to correct R2 for longer fragments and both R1/2 for shorter ones. Results: We quantified the prevalence of such errors at JEs of double-stranded cfDNA molecules in plasma of 2 healthy subjects. We compared library preparations on single- (SPLAT-Seq) vs double-stranded DNA (EM-Seq), and strikingly found ~40% of CpGs to be involved in jagged cfDNA ends and impacted by misrepair during enzymatic ER of dsDNA. This resulted in an artifactual 13.1-16.3% lower global methylation across the genome, an effect that was exaggerated in hypermethylated (hyper) regions (~69% of genome). While ER-associated misrepair resulted in ~10.2% lower CpG methylation within the first 110bp of cfDNA fragments (R1 AUC=0.95, R2 AUC=0.81), when comparing R1/R2, the final 30bp showed even larger biases (13.8-77.4%). Remarkably, JEEPERS could accurately detect and correct these biases and produce highly uniform R1/R2 profiles in EM-Seq on par with SPLAT-Seq, even when considering short cfDNA fragments (<141bp) with substantial JE fractions. We also compared JEEPERS performance with other methods, including simple JE trimming or reliance on R1 alone, finding superior performance by JEEPERS for methylation quantification. Conclusions: ER distorts a large fraction of CpGs across the genome clustering at cfDNA fragment ends, resulting in substantially decreased methylation levels in cfDNA. JEEPERS accurately recovers methylation levels distorted by such misrepair. This not only enables more accurate methylation levels, but also retains cfDNA fragmentation features and duplex information. These advantages allow not only superior cfDNA genotyping of germline and somatic variants, but also enable better identification of allele-specific methylation, hemi-methylation, and gene expression inferences, as relevant for cancer detection. Citation Format: Rui Wang, Emily G. Hamilton, Angela Hui, Diego Almanza, Mohammad S. Esfahani, Maximilian Diehn, Ash A. Alizadeh. Improved cfDNA methylation profiling through correction of misrepaired jagged-ends [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 1024.

Double‐blind peer review affects reviewer ratings and editor decisions at an ecology journal

Abstract There is substantial evidence that systemic biases influence the scholarly peer review process. Many scholars have advocated for double‐blind peer review (also known as double‐anonymous review) to reduce these biases. However, the effectiveness of double‐blind peer review in eliminating biases is uncertain because few randomized trials have manipulated blinding of author identities for journal submissions and those that have are generally small or provide few insights on how it influences reviewer biases. In 2019, Functional Ecology began a large, randomized trial, using real manuscript submissions, to evaluate the various consequences of shifting to double‐blind peer review. Research papers submitted to the journal were randomly assigned to be reviewed with author identities blinded to reviewers (double‐blind review) or with authors identified to reviewers (single‐blind review). In this paper, we explore the effect of blinding on the outcomes of peer review, examining reviewer ratings and editorial decisions, and ask whether author gender and/or location mediate the effects of review type. Double‐blind review reduced the average success of manuscripts in peer review; papers reviewed with author identities blinded received on average lower ratings from reviewers and were less likely to be invited for revision or resubmission. However, the effect of review treatment varied with the author's location. Papers with first authors residing in countries with a higher human development index (HDI) and/or higher average English proficiency fared much better than those from countries with a lower HDI and lower English proficiency, but only when author identities were known to reviewers; outcomes were similar between demographic groups when author identities were not known to reviewers. Blinding author identities had no effect on gender differences in reviewer ratings or editor decisions. Our data provide strong evidence that authors from higher income and/or English‐speaking countries receive significant benefits (a large positive bias) to being identified to reviewers during the peer review process and that anonymizing author‐identities (e.g. double‐blind review) reduces this bias, making the peer review process more equitable. We suggest that offering optional blinding of author identities, as some journals allow, is unlikely to substantially reduce the biases that exist because authors from higher‐income and English‐speaking countries are the least likely to choose to be reviewed with their identity anonymized. Read the free Plain Language Summary for this article on the Journal blog.

Agreement of Temperatures Measured Using a Non-Contact Infrared Thermometer With a Rectal Digital Thermometer in Horses.

Evaluating the body temperature of horses is an essential tool for monitoring horse health and biosecurity in groups of horses. Temperatures of horses and foals are determined most often using rectal thermometry. Rectal thermometry has limitations that include safety considerations for horses and humans. Thus, we investigated the agreement between a noncontact infrared thermometer (NCIT) and a rectal digital thermometer in 142 horses and 34 foals. For each horse and foal, measurements using the NCIT were collected from the forehead (n=2) or neck (n=1) and with a rectal digital thermometer (n=1). Although the NCIT demonstrated good reliability (i.e. repeatability of measurements), a large negative bias (nearly 2°F (-16.7°C) in adult horses and >3°F (-16.1°C) in foals) was observed between readings from the NCIT and the rectal thermometer in healthy horses. Although horses with febrile illness were not included in the study, our results indicate that the large and inconsistent bias observed with the NCIT indicates that these devices will not be a suitable substitute for rectal thermometry for obtaining valid estimates of core body temperature in horses.

Controlling for cohort effects in accelerated longitudinal designs using continuous- and discrete-time dynamic models.

Accelerated longitudinal designs (ALDs) allow examining developmental changes over a period of time longer than the duration of the study. In ALDs, participants enter the study at different ages (i.e., different cohorts), and provide measures during a time frame shorter than the total study. They key assumption is that participants from the different cohorts come from the same population and, therefore, can be assumed to share the same general trajectory. The consequences of not meeting that assumption have not been examined systematically. In this article, we propose an approach to detect and control for cohort differences in ALDs using latent change score models in both discrete and continuous time. We evaluated the effectiveness of such a method through a Monte Carlo study. Our results indicate that, in a broad set of empirically relevant conditions, both latent change score (LCS) specifications can adequately estimate cohort effects ranging from very small to very large, with slightly better performance of the continuous-time version. Across all conditions, cohort effects on the asymptotic level (dAs) caused much larger bias than on the latent initial level (d₀). When cohort differences were present, including them in the model led to unbiased estimates. In contrast, not including them led to tenable results only when such differences were not large (d₀ ≤ 1 and dAs ≤ .2). Among the sampling schedules evaluated, those including at least three measurements per participant over 4 years or more led to the best performance. Based on our findings, we offer recommendations regarding study designs and data analysis. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

Harmonization of cognitive screening tools for dementia across diverse samples: A simulation study.

Research focusing on cognitive aging and dementia is a global endeavor. However, cross-national differences in cognition are embedded in other sociocultural differences, precluding direct comparisons of test scores. Such comparisons can be facilitated by co-calibration using item response theory (IRT). The goal of this study was to explore, using simulation, the necessary conditions for accurate harmonization of cognitive data. Neuropsychological test scores from the US Health and Retirement Study (HRS) and the Mexican Health and Aging Study (MHAS) were subjected to IRT analysis to estimate item parameters and sample means and standard deviations. These estimates were used to generate simulated item response patterns under 10 scenarios that adjusted the quality and quantity of linking items used in harmonization. IRT-derived factor scores were compared to the known population values to assess bias, efficiency, accuracy, and reliability of the harmonized data. The current configuration of HRS and MHAS data was not suitable for harmonization, as poor linking item quality led to large bias in both cohorts. Scenarios with more numerous and higher quality linking items led to less biased and more accurate harmonization. Linking items must possess low measurement error across the range of latent ability for co-calibration to be successful. We developed a statistical simulation platform to evaluate the degree to which cross-sample harmonization accuracy varies as a function of the quality and quantity of linking items.Two large studies of aging-one in Mexico and one in the United States-use three common items to measure cognition.These three common items have weak correspondence with the ability being measured and are all low in difficulty.Harmonized scores derived from the three common linking items will provide biased and inaccurate estimates of cognitive ability.Harmonization accuracy is greatest when linking items vary in difficulty and are strongly related to the ability being measured.

Daytime Sleep-Tracking Performance of Four Commercial Wearable Devices During Unrestricted Home Sleep.

Previous studies have found that many commercial wearable devices can accurately track sleep-wake patterns in laboratory or home settings. However, nearly all previous studies tested devices under conditions with fixed time in bed (TIB) and during nighttime sleep episodes only. Despite its relevance to shift workers and others with irregular sleep schedules, it is largely unknown how devices track daytime sleep. Therefore, we tested the sleep-tracking performance of four commercial wearable devices during unrestricted home daytime sleep. Participants were 16 healthy young adults (6 men, 10 women; 26.6 ± 4.6 years, mean ± SD) with habitual daytime sleep schedules. Participants slept at home for 1 week under unrestricted conditions (ie, self-selecting TIB) using a set of four commercial wearable devices and completed reference sleep logs. Wearables included the Fatigue Science ReadiBand, Fitbit Inspire HR, Oura Ring, and Polar Vantage V Titan. Daytime sleep episode TIB biases and frequencies of missed and false-positive daytime sleep episodes were examined. TIB bias was low in general for all devices on most daytime sleep episodes, but some exhibited large biases (eg, >1 h). Total missed daytime sleep episodes were as follows: Fatigue Science: 3.6%; Fitbit: 4.8%; Oura: 6.0%; Polar: 37.3%. Missed episodes occurred most often when TIB was short (eg, naps <4 h). When daytime sleep episodes were recorded, the devices generally exhibited similar performance for tracking TIB (ie, most episodes had low bias). However, the devices failed to detect some daytime episodes, which occurred most often when TIB was short, but varied across devices (especially Polar, which missed over one-third of episodes). Findings suggest that accurate daytime sleep tracking is largely achievable with commercial wearable devices. However, performance differences for missed recordings suggest that some devices vary in reliability (especially for naps), but improvements could likely be made with changes to algorithm sensitivities.

New Methodology for Unbiased Ground-Motion Intensity Conversion Equations

ABSTRACT Ground-motion intensity conversion equations (GMICEs) allow for conversions between ground-motion amplitude and shaking intensity. The current methods used to develop GMICEs model the dependence of the intensity on the peak ground-motion (PGM) parameter. For some models, there is a second step that models the magnitude and distance dependence of the residuals from the initial regression. We show that this approach for developing GMICEs works well for estimating the intensity from median ground motions, but for ground-motion values away from the medians, the intensities estimated by the GMICE can have large bias, with overprediction for positive PGM residuals and underprediction for negative PGM residuals. The bias is due to an implicit assumption in the current approach that there is a direct causal relation between intensity and the ground-motion parameter and that the residuals of the intensity and ground-motion parameter are fully correlated. We present two alternative methodologies for developing GMICEs that do not suffer from this bias. The first method includes the magnitude and distance scaling of the GMICE in the initial regression for the scaling with the PGM. The second method excludes the magnitude and distance terms but includes the number of standard deviations of the PGM (ϵ) as an additional parameter in the GMICE. Using a synthetic data set of intensity and peak ground acceleration values, we show that the GMICE developed using the proposed method leads to more accurate estimates of the intensity than current methods. We also discuss implications of using GMICEs based on the proposed methods for the evaluation of probabilistic hazard maps and as input to ShakeAlert estimates.

Nonlinear Bias Correction of the FY-4A AGRI Infrared Radiance Data Based on the Random Forest

Bias correction is a key prerequisite for radiance data assimilation. Directly assimilating the radiance observations generally involves large systematic biases affecting the numerical prediction accuracy. In this study, a nonlinear bias correction scheme with Random Forest (RF) technology is firstly proposed based on the Fengyun-4A (FY-4A) Advanced Geosynchronous Radiation Imager (AGRI) channels 9–10 observations in the Weather Research and Forecasting Data Assimilation (WRFDA) system. Two different settings of the predictors are additionally designed and evaluated based on the performance of the RF model. It seems that an apparent scene temperature-dependent bias could be effectively resolved by the RF scheme when applying the RF method with newly added predictors. Results suggest that the proposed nonlinear scheme of RF performs better than the linear scheme does in terms of reducing the systematic biases. A more idealized error distribution of observation minus background (OMB) is found in the RF-based experiments that measure the nonlinear relationship between the OMB biases and the predictors when using the Gaussian distribution as the reference. Furthermore, the RF scheme shows a consistent improvement in bias correction with the potential to ameliorate the atmospheric variables of analyses.

Surface Modification of a Titanium Carbide MXene Memristor to Enhance Memory Window and Low‐Power Operation

AbstractWith the demand for low‐power‐operating artificial intelligence systems, bio‐inspired memristor devices exhibit potential in terms of high‐density memory functions and the emulation of the synaptic dynamics of the human brain. The 2D material MXene attracts considerable interest for use in resistive‐switching memory and artificial synapse devices owing to its excellent physicochemical properties in memristor devices. However, few memristive and synaptic MXene devices that display increased switching performances are reported, with no significant results. Herein, the conductivity of MXene (Ti3C2Tx) is engineered via etching and oxidation to enhance the switching performance of the device. The exceptional properties of partially oxidized MXene memristors include large memory windows and low threshold biases, and the complex spike‐timing‐dependent plasticity synaptic rules are also emulated. The low threshold potential distribution, reliable retention time (104 s), and distinct resistance states with a high ON–OFF ratio (&gt;104) are the main memory‐related features of this device. The experimentally determined switching potentials of the optimized device are also uniformly distributed, according to a statistical probability‐based approach. This investigation may promote the essential material properties for use in high‐density non‐volatile memory storage and artificial synapse systems in the field of innovative nanoelectronic devices.

Seasonal dependent suitability of physical parameterizations to simulate precipitation over the Himalayan headwater

The Himalayan ecosystem is fragile and needs robust management strategies for sustainability of natural resources such as water and vegetation. Therefore, reliable precipitation estimation becomes quite important from operational and regulation standpoints. It is crucial for numerous activities including policy/planning, agriculture, reservoir operations, disaster management, and others. In addition, reliable information on temporal variability of precipitation is also crucial for various applications such as agricultural and hydrological. The western Himalaya receives two distinct weather systems during summer and winter. Summer is responsible (largely) for rainfall and winter is for snowfall. Therefore, we hypothesize that there may not be a single set of parameterization schemes that can represent well both the weather systems. To investigate, we set up the WRF modeling system and performed six experiments with a combination of three microphysics (MP3, MP3, and WSM6) and two cumulus schemes (KF, and BMJ). It was found that the precipitation along the Himalayan foothills (near to basin terminal) is underestimated in four out of six experiments. Only experiments with BMJ cumulus scheme along with WSM6 and MP8 microphysics were able to show a considerable amount of precipitation along these foothills. It was noted that all six experiments showed high precipitation in the upstream region and over the mountain peaks and ridges in North-Western Himalaya. For DJF, each experiment was found to have large biases and none of them represented the observation with high confidence. However, the selection of observation reference data itself is a challenging task because of data paucity in this region. Therefore, the closest experiment to the most appropriate observation was selected as the reliable configuration (MP8_KF: MP8 microphysics and KF cumulus scheme) for DJF precipitation simulation. In this study we have, for the first time, reported the role of seasonal sensitivity for the climate scale simulations as we found that different schemes were suitable for different weather systems.

Degenerate optical parametric amplification in CMOS silicon

Silicon is a common material for photonics due to its favorable optical properties in the telecom and mid-wave IR bands, as well as compatibility with a wide range of complementary metal–oxide semiconductor (CMOS) foundry processes. Crystalline inversion symmetry precludes silicon from natively exhibiting second-order nonlinear optical processes. In this work, we build on recent works in silicon photonics that break this material symmetry using large bias fields, thereby enabling χ(2) interactions. Using this approach, we demonstrate both second-harmonic generation (with a normalized efficiency of 0.20%W−1cm−2) and, to our knowledge, the first degenerate χ(2) optical parametric amplifier (with an estimated normalized gain of 0.6dBW−1/2cm−1) using silicon-on-insulator waveguides fabricated in a CMOS-compatible commercial foundry. We expect this technology to enable the integration of novel nonlinear optical devices such as optical parametric amplifiers, oscillators, and frequency converters into large-scale, hybrid photonic–electronic systems by leveraging the extensive ecosystem of CMOS fabrication.

Strong lateral exchange coupling and current-induced switching in single-layer ferrimagnetic films with patterned compensation temperature

Interlayer couplings are widely used in spintronic devices based on magnetic multilayers. Here, the authors demonstrate instead a strong tunable coupling between adjacent regions of a single ferrimagnetic layer, based on the exchange interaction. They unveil a large lateral exchange bias by spatially patterning the compensation temperature of the ferrimagnet. Furthermore, they show electrical switching of ferrimagnetic domains utilizing the lateral exchange coupling mechanism in combination with spin-orbit torques.

Deep Learning for Improved Precision and Reproducibility of Left Ventricular Strain in Echocardiography: A Test-Retest Study

Assessment of left ventricular (LV) function by echocardiography is hampered by modest test-retest reproducibility. A novel artificial intelligence (AI) method based on deep learning provides fully automated measurements of LV global longitudinal strain (GLS) and may improve the clinical utility of echocardiography by reducing user-related variability. The aim of this study was to assess within-patient test-retest reproducibility of LV GLS measured by the novel AI method in repeated echocardiograms recorded by different echocardiographers and to compare the results to manual measurements. Two test-retest data sets (n=40 and n=32) were obtained at separate centers. Repeated recordings were acquired in immediate succession by 2 different echocardiographers at each center. For each data set, 4 readers measured GLS in both recordings using a semiautomatic method to construct test-retest interreader and intrareader scenarios. Agreement, mean absolute difference, and minimal detectable change (MDC) were compared to analyses by AI. In a subset of 10 patients, beat-to-beat variability in 3 cardiac cycles was assessed by 2 readers and AI. Test-retest variability was lower with AI compared with interreader scenarios (data set I: MDC= 3.7 vs 5.5, mean absolute difference= 1.4 vs 2.1, respectively; data set II: MDC= 3.9 vs 5.2, mean absolute difference= 1.6 vs 1.9, respectively; all P<.05). There was bias in GLS measurements in 13 of 24 test-retest interreader scenarios (largest bias, 3.2 strain units). In contrast, there was no bias in measurements by AI. Beat-to-beat MDCs were 1,5, 2.1, and 2.3 for AI and the 2 readers, respectively. Processing time for analyses of GLS by the AI method was 7.9±2.8seconds. A fast AI method for automated measurements of LV GLS reduced test-retest variability and removed bias between readers in both test-retest data sets. By improving the precision and reproducibility, AI may increase the clinical utility of echocardiography.

Exploring the reliability of 222Rn as a tracer of groundwater age in alluvial aquifers: Insights from the explicit simulation of variable 222Rn production

Knowledge of groundwater residence times (GRT; the time elapsed since surface water infiltration) between losing rivers and pumping wells is crucial for management of water resources in alluvial aquifers. The radioactive noble gas radon-222 (222Rn) has been used for decades as a natural indicator of surface water infiltration, as it can provide quantitative information on GRT. However, models using 222Rn as a tracer of GRT are often based on a set of highly simplifying assumptions, including spatially homogenous 222Rn production and exclusively advective mass transport within the aquifer. In this paper, we use the integrated surface-subsurface hydrological model HydroGeoSphere (HGS) to simulate 222Rn transport, production, and decay in a bank filtration context. Spatially variable 222Rn production, based on experimental data, is explicitly considered. We show that variable 222Rn production rates, coupled with hydrodispersive mixing of groundwater, may lead to large biases in GRT estimates. Under certain transient conditions however, changes in tracer-derived GRTs correlate well with changes in mean groundwater age. Whereas 222Rn-derived GRTs may only be reliable under a narrow range of field conditions, 222Rn may serve as a powerful tracer of changes in mean GRT even in complex and heterogenous environments.

Urban environmental inequalities in Latin America: A scoping review

Latin America is among the most urbanized regions in the world. However, cities in this region usually present an uneven distribution of environmental amenities and hazards between socioeconomic sectors, translating into prevalent issues of environmental inequalities. Urban environmental inequalities have been widely covered by studies from developed regions, but less attention has been placed on highly urbanized regions from the developing world. In this work, we review the topic of urban environmental inequalities in Latin America, aiming to identify the temporal and geographical trends of published research, and revise the main themes framing the research on environmental inequalities in this region. We searched for literature in three databases (i.e. Scielo, Scopus, Web of Science) using three different languages (i.e. English, Spanish, Portuguese). We identified a total of 57 articles published on the topic. These articles show an increasing rate of publication and citations during the last decade, with increasing proportion of papers published in English in journals covered by international databases. There is a large geographical bias towards a few countries and cities, and more than half of Latin-American countries have no research on this topic yet. While in North America race and ethnicity have been often associated to urban environmental inequalities, in Latin America most of studies relate environmental inequalities to social and economic variables. These results highlight the relevance of carrying out this type of research in developing regions, as these may provide key novel insights for generating better solutions to address urban environmental inequalities.

Meta-Analysis of 49 Roche Oncology Trials Comparing Blinded Independent Central Review (BICR) and Local Evaluation to Assess the Value of BICR.

Blinded independent central review (BICR) of radiographic images is frequently conducted in oncology trials to address the potential bias of local evaluation (LE) of endpoints such as progression-free survival (PFS) and objective response rate (ORR). Given that BICR is a complex and costly process, we evaluated the agreement between LE- and BICR-based treatment effect results and the impact of BICR on regulatory decision-making. Meta-analyses were performed using hazard ratios (HRs) for PFS and odds ratios (ORs) for ORR from all randomized Roche-supported oncology clinical trials during 2006-2020 that had both LE and BICR results (49 studies with a total of over 32 000 patients). Overall, the evaluation bias of LE overestimating the treatment effect compared with BICR based on PFS was numerically small and not clinically meaningful, especially for double-blind studies (HR ratio between BICR and LE: 1.044). A larger bias is more likely to occur in studies with open-label design, smaller sample sizes, or an unequal randomization ratio. The majority (87%) of the PFS comparisons led to the same statistical inference by BICR and LE. For ORR, a high degree of agreement between BICR and LE results was also observed (OR ratio of 1.065), although the agreement was slightly lower than for PFS. BICR did not notably impact the study interpretation nor drive the sponsor's regulatory submission decisions. Hence, if bias can be diminished by appropriate means, LE is deemed as reliable as BICR for certain study settings.

A self-attention-based neural network for three-dimensional multivariate modeling and its skillful ENSO predictions.

Large biases and uncertainties remain in real-time predictions of El Niño-Southern Oscillation (ENSO) using process-based dynamical models; recent advances in data-driven deep learning algorithms provide a promising mean to achieve superior skill in the tropical Pacific sea surface temperature (SST) modeling. Here, a specific self-attention-based neural network model is developed for ENSO predictions based on the much sought-after Transformer model, named 3D-Geoformer, which is used to predict three-dimensional (3D) upper-ocean temperature anomalies and wind stress anomalies. This purely data-driven and time-space attention-enhanced model achieves surprisingly high correlation skills for Niño 3.4 SST anomaly predictions made 18 months in advance and initiated beginning in boreal spring. Further, sensitivity experiments demonstrate that the 3D-Geoformer model can depict the evolution of upper-ocean temperature and the coupled ocean-atmosphere dynamics following the Bjerknes feedback mechanism during ENSO cycles. Such successful realizations of the self-attention-based model in ENSO predictions indicate its great potential for multidimensional spatiotemporal modeling in geoscience.

Spin filters based on two-dimensional materials Co2Si and Cu2Si

Spintronic devices have several advantages compared with conventional electronic devices, including non-volatility, faster data processing speed, higher integration densities, less electric power consumption and so on. However, we still face challenges for efficiently generating and injecting pure spin polarized current. In this work, we utilize two kinds of two-dimensional materials Co2Si and Cu2Si with both lattice match and band match to construct devices and then research their spin filter efficiency. The spin filter efficiency can be improved effectively either by an appropriate gate voltage at Co2Si region, or by series connection. In both cases the filter efficiencies are much larger than two-dimensional prepared Fe3GeTe2 spin valve and ferromagnetic metallic chairlike O-graphene-H. Also at a quite small bias, we obtain a comparable spin polarized current as those obtained in Fe3GeTe2 spin valve and O-graphene-H obtained at a much larger bias.

Imprecision remains to be improved in the measurement of serum cystatin C with heterogeneous systems.

Except for the large bias of some measurement systems for serum cystatin C (CysC) measurements, unacceptable imprecision has been observed for the heterogenous system. This study analyzed the external quality assessment (EQA) results in 2018-2021 to provide an insight into the imprecision of CysC assays. Five EQA samples were sent to participating laboratories every year. Participants were divided into reagent/calibrator-based peer groups, for which the robust mean of each sample and robust coefficient of variation (CV) were calculated by Algorithm A from ISO 13528. Peers with more than 12 participants per year were selected for further analysis. The limit of CV was determined to be 4.85% based on clinical application requirements. The concentration-related effect on CVs was investigated using logarithmic curve fitting; the difference in medians and robust CVs between instrument-based subgroups was also evaluated. The total number of participating laboratories increased from 845 to 1,695 in four years and heterogeneous systems remained the mainstream (≥85%). Of 18 peers with≥12 participants, those using homogeneous systems showed relatively steady and small CVs over four years, with the mean four-year CVs ranging from 3.21 to 3.68%. Some peers using heterogenous systems showed reduced CVs over four years, while 7/15 still had unacceptable CVs in 2021 (5.01-8.34%). Six peers showed larger CVs at the low or high concentrations, and some instrument-based subgroups presented greater imprecision than others. More efforts should be made to improve theimprecision of heterogeneous systems for CysC measurement.