High Relative Errors Research Articles

Applicability of NMR spectroscopy to quantify microplastics across varying concentrations in polymer mixtures.

Quantitative nuclear magnetic resonance (qNMR) spectroscopy could potentially be used for environmental microplastic analyses, provided the challenges posed by mixed polymer samples with varying concentrations and overlapping signals are understood. This study investigates the feasibility of qNMR as a reliable and cost-efficient method for quantifying synthetic polymers in mixtures of low and varying concentrations, addressing key challenges and limitations. Polymer mixtures were analysed using deuterated chloroform (CDCl3) and deuterated tetrahydrofuran (THF-d8) as solvents, with polystyrene (PS), polybutadiene-cis (PB), polyisoprene-cis (PI), polyvinyl chloride (PVC), polyurethane (PU), and polylactic acid (PLA) as selected polymers. Mixtures contained either low and high concentrations of each polymer or equal concentrations of all six polymers. Polymer concentrations were measured using the internal standard method. The method showed low relative errors for low concentrations of PS in CDCl3 and PVC in THF-d8, with values of -5% and 0%, respectively, while PB and PI in CDCl3 show relative errors of +5% and -3%, respectively. We observe significant linearity between nominal and measured concentrations with R 2 values ranging from 0.9655 to 0.9981, except for PU, which had high relative errors and poor linearity (R 2 = 0.9548). Moreover, simultaneous quantification of six polymers in THF-d8 proves effective at intermediate concentrations. However, overlapping proton signals are observed, causing high-concentration polymers to mask low-concentration ones. While this study demonstrates low limit of quantification (LOQ) and advances in simultaneous polymer quantification, further research is needed to improve qNMR accuracy for mixed polymer samples and environmentally relevant concentrations.

Accelerometers can correctly count orthopaedic patients' early post-operative steps while using walking aids.

Effective rehabilitation after orthopaedic surgery is critical. The early post-operative phase is increasingly managed in outpatient settings, necessitating objective measures such as step counts to monitor rehabilitation progress. However, it remains unclear if commercially available wearables or accelerometers using simple algorithms can accurately count steps in early post-operative conditions. We hypothesised that only accelerometers could accurately determine the number of steps under these conditions. This case series involved 20 healthy subjects, 7 female and 13 males, walking in a circle at varying speeds under partial loading with three different walking aids (forearm crutches, walking frame and rolling walker) and four wearables (Vivofit 4, Fenix 3HR, Fitbit Charge 3 and Omron HJ-325) and one accelerometer (AX6) worn on the wrist, hip and ankle. The two-point and modified three-point gait patterns commonly used post-operatively were simulated. The primary end point was the relative error (RE), defined as RE = (manual count - automated count)/manual count, of each wearable measurement compared to visual and video step counting, the gold standard. The RE of AX6 and Fitbit was less than 0.1 for all walking aids except the rolling walker, with AX6 showing the lowest standard deviation (SD) compared to other wearables. Other wearables had significantly higher RE. Increased gait speed generally improved accuracy, reducing RE in most devices, except for the AX6, which showed the opposite trend. At 0.6 m/s, only AX6 achieved an RE below 0.1. The ankle was identified as the best measuring location. During the early post-operative period, commercial wearables can only accurately count steps under specific conditions and should be used cautiously for monitoring steps in the early post-operative phase. However, accelerometers with appropriate coding appear suitable for this purpose. Level III diagnostic study.

Data-driven modeling approaches for pressure drop prediction in a multi-phase flow system

Accurate prediction of pressure drops in multi-phase flow systems is essential for optimizing processes in industries such as oil and gas, where operational efficiency and safety depend on reliable modeling. Traditional models often need help with the complexities of multi-phase flow dynamics, resulting in high relative errors, particularly under varying flow regimes. In this study, we simulate a comprehensive multiphase flow experimental data collected from the lab. This study presents innovative methods for accurately modeling pressure drops in multi-phase flow systems. It also studies the complicated dynamics of multi-phase flows, which are flows with more than one phase at the same time. It does this by using two different data-driven models, nonlinear ARX and Hammerstein-Wiener, instead of neural networks (NNs), so that the models don’t get too good at fitting environments with lots of changes and little data. Our research applies system identification approaches to the intricacies of this domain, providing new insights into choosing the best appropriate modeling strategy for multi phase flow systems, taking into account their distinct properties. The experimental results show that the nonlinear Hammerstein-Wiener and ARX models were much better than other methods, with fitting accuracy rates of 81.12% for the Hammerstein-Wiener model and 86.52% for the ARX model. This study helps the creation of more advanced control algorithms by providing a reliable way to guess when the pressure drops and showing how to choose a model that fits the properties of the multi-phase flow. These findings contribute to enhanced pressure management and optimization strategies, setting a foundation for future studies on real-time flow control and broader industrial applications.

Assessment of the impact of NO2 contribution on aerosol-optical-depth measurements at several sites worldwide

Abstract. This work aims at investigating the effect of NO2 absorption on aerosol-optical-depth (AOD) measurements and Ångström exponent (AE) retrievals of sun photometers by the synergistic use of accurate NO2 characterization for optical-depth estimation from co-located ground-based measurements. The analysis was performed for ∼ 7 years (2017–2023) at several sites worldwide for the AOD measurements and AE retrievals by Aerosol Robotic Network (AERONET) sun photometers which use OMI (Ozone Monitoring Instrument) climatology for NO2 representation. The differences in AOD and AE retrievals by NO2 absorption are accounted for using high-frequency columnar NO2 measurements by a co-located Pandora spectroradiometer belonging to the Pandonia Global Network (PGN). NO2 absorption affects the AOD measurements in UV-Vis (visible) range, and we found that the AOD bias is the most affected at 380 nm by NO2 differences, followed by 440, 340, and 500 nm, respectively. AERONET AOD was found to be overestimated in half of the cases, while also underestimated in other cases as an impact of the NO2 difference from “real” (PGN NO2) values. Overestimations or underestimations are relatively low. About one-third of these stations showed a mean difference in NO2 and AOD (at 380 and 440 nm) above 0.5 × 10−4 mol m−2 and 0.002, respectively, which can be considered a systematic contribution to the uncertainties in the AOD measurements that are reported to be of the order of 0.01. However, under extreme NO2 loading scenarios (i.e. 10 % highest differences) at highly urbanized/industrialized locations, even higher AOD differences were observed that were at the limit of or higher than the reported 0.01 uncertainty in the AOD measurement. PGN NO2-based sensitivity analysis of AOD difference suggested that for PGN NO2 varying between 2 × 10−4 and 8 × 10−4 mol m−2, the median AOD differences were found to rise above 0.01 (even above 0.02) with the increase in NO2 threshold (i.e. the lower limit from 2 × 10−4 to 8 × 10−4 mol m−2). The AOD-derivative product, AE, was also affected by the NO2 correction (discrepancies between the AERONET OMI climatological representation of NO2 values and the real PGN NO2 measurements) on the spectral AOD. Normalized frequency distribution of AE (at 440–870 and 340–440 nm wavelength pair) was found to be narrower for a broader AOD distribution for some stations, and vice versa for other stations, and a higher relative error at the shorter wavelength (among the wavelength pairs used for AE estimation) led to a shift in the peak of the AE difference distribution towards a higher positive value, while a higher relative error at a lower wavelength shifted the AE difference distribution to a negative value for the AOD overestimation case, and vice versa for the AOD underestimation case. For rural locations, the mean NO2 differences were found to be mostly below 0.50 × 10−4 mol m−2, with the corresponding AOD differences being below 0.002, and in extreme NO2 loading scenarios, it went above this value and reached above 1.00 × 10−4 mol m−2 for some stations, leading to higher AOD differences but below 0.005. Finally, AOD and AE trends were calculated based on the original AERONET AOD (based on AERONET OMI climatological NO2), and its comparison with the mean differences in the AERONET and PGN NO2-corrected AOD was indicative of how NO2 correction could potentially affect realistic AOD trends.

Effect of Dynamic Balance Training on Physical Function and Proprioception in Individuals with Knee Osteoarthritis

Background: Osteoarthritis (OA) is a prevalent joint disorder affecting synovial joints, particularly the knee. It leads to degeneration of cartilage, causing pain and reduced physical function. Effective treatment strategies are essential for managing symptoms and improving quality of life. Objective: To determine the effect of dynamic balance training on physical function and proprioception in individuals with knee osteoarthritis. Methods: This quasi-experimental trial was conducted at Riphah International University Lahore, Aziz Medical Clinic, and Ali Hospital, Mughal Pura Lahore. A total of 48 patients with anterior knee pain were recruited and divided into two groups: one receiving dynamic balance training and the other receiving conventional physical therapy (control). Non-probability convenience sampling was used. Outcomes were measured using NPRS, a goniometer, the Community Balance and Mobility Scale (CB&M), and the Western Ontario and McMaster Universities Arthritis Index (WOMAC). Knee proprioception was assessed by joint position sense using a universal goniometer, with the average of three angles taken for the final reading. Data were analyzed using SPSS version 25. Results: Significant differences were observed in joint proprioception between weight-bearing (WB) and non-weight-bearing (NWB) procedures. The mean difference for WB was -3.07 compared to -2.18 for NWB, with a higher relative error in WB (p < 0.05). Dynamic balance training resulted in greater reductions in NPRS scores and improvements in WOMAC scores compared to conventional physical therapy (p < 0.05). Conclusion: Dynamic balance training was more effective than conventional physical therapy in improving physical function and proprioception in individuals with knee osteoarthritis.

Improved filtering methods to suppress cardiovascular contamination in electrical impedance tomography recordings

Objective. Electrical impedance tomography (EIT) produces clinical useful visualization of the distribution of ventilation inside the lungs. The accuracy of EIT-derived parameters can be compromised by the cardiovascular signal. Removal of these artefacts is challenging due to spectral overlapping of the ventilatory and cardiovascular signal components and their time-varying frequencies. We designed and evaluated advanced filtering techniques and hypothesized that these would outperform traditional low-pass filters. Approach. Three filter techniques were developed and compared against traditional low-pass filtering: multiple digital notch filtering (MDN), empirical mode decomposition (EMD) and the maximal overlap discrete wavelet transform (MODWT). The performance of the filtering techniques was evaluated (1) in the time domain (2) in the frequency domain (3) by visual inspection. We evaluated the performance using simulated contaminated EIT data and data from 15 adult and neonatal intensive care unit patients. Main result. Each filter technique exhibited varying degrees of effectiveness and limitations. Quality measures in the time domain showed the best performance for MDN filtering. The signal to noise ratio was best for DLP, but at the cost of a high relative and removal error. MDN outbalanced the performance resulting in a good SNR with a low relative and removal error. MDN, EMD and MODWT performed similar in the frequency domain and were successful in removing the high frequency components of the data. Significance. Advanced filtering techniques have benefits compared to traditional filters but are not always better. MDN filtering outperformed EMD and MODWT regarding quality measures in the time domain. This study emphasizes the need for careful consideration when choosing a filtering approach, depending on the dataset and the clinical/research question.

Improved engineering model for water absorption in softwoods

Modeling of moisture transfer in porous materials is typically partitioned into liquid transport and vapor diffusion. Coefficients for liquid transport are often derived from laboratory measurements of water absorption by partial immersion. Although the assumption that capillary liquid transport is dominant while vapor diffusion is negligible is a reasonable assumption for mineral-based materials, recent work indicates that it is not representative of water absorption in wood across the grain. This article builds on earlier work and develops an improved engineering model for liquid water transport in softwoods. The approach uses a constant liquid diffusivity value, relies primarily on ordinary hygrothermal property measurements, and partitions vapor and liquid transport in a practical way that is informed by current understanding of wood–water interactions. The model is optimized using one-dimensional simulations calibrated against measured water uptake across a range of softwood species. Laboratory measurements were conducted to investigate key properties of eastern white cedar, which had the highest relative error in a previous version of the model, and to examine the practicality and relevance of vacuum saturation and free saturation for southern yellow pine. The model accuracy is improved by paying closer attention to the moisture storage function and capillary saturation value.

A Bidirectional Deep Learning Approach for Designing MEMS Sensors

To achieve the desired characteristics for MEMS sensors, traditional design process obtains the geometrical parameters based on complex theoretical calculations and interactive finite element (FE) simulations, which are time-consuming and data-consuming. To solve the above problems, a data-driven bidirectional design approach based on deep learning (DL) method is introduced to improve the design efficiency of MEMS sensors in this work. By using the piezoresistive acceleration sensor as a design example, the forward artificial neural network (ANN) with the sensor geometrical parameters as the input and the sensor performance as the output is trained and realized by using 1000 groups of data collected through FE simulation. This forward ANN can accurately predict the sensor performance including the measurement range, sensitivity, and resonant frequency. In addition, the inverse ANN with the sensor performance as the input and the sensor geometrical parameters as the output is also achieved by using a tandem network. This inverse ANN can provide the geometrical parameters directly and instantly according to the target performance. Both the forward and inverse networks cost only about 6 ms for each task and the mean relative errors are less than 3%. The high efficiency and low relative error indicate that DL is a promising approach to improve the design efficiency for MEMS sensors.

Computation of the Distribution of the Sum of Independent Negative Binomial Random Variables

The distribution of the sum of negative binomial random variables has a special role in insurance mathematics, actuarial sciences, and ecology. Two methods to estimate this distribution have been published: a finite-sum exact expression and a series expression by convolution. We compare both methods, as well as a new normalized saddlepoint approximation, and normal and single distribution negative binomial approximations. We show that the exact series expression used lots of memory when the number of random variables was high (>7). The normalized saddlepoint approximation gives an output with a high relative error (around 3–5%), which can be a problem in some situations. The convolution method is a good compromise for applied practitioners, considering the amount of memory used, the computing time, and the precision of the estimates. However, a simplistic implementation of the algorithm could produce incorrect results due to the non-monotony of the convergence rate. The tolerance limit must be chosen depending on the expected magnitude order of the estimate, for which we used the answer generated by the saddlepoint approximation. Finally, the normal and negative binomial approximations should not be used, as they produced outputs with a very low accuracy.

A General Framework for Automated Accurate Calculation of b-Matrix (Auto-b) in Diffusion MRI Pulse Sequences

To derive accurate diffusion metrics, both imaging and diffusion-sensitizing gradient pulses should be accounted for when calculating the diffusion-weighted b-matrix. However, it is complex to derive analytical solutions due to complicated interactions between gradient pulses, including orthogonal directions. This study proposes a general framework to calculate the b-matrix automatically (dubbed as Auto-b). Based on the divide-and-conquer approach, the b-matrix calculation is appropriately segmented, and the symbolic mathematical library is applied to handle integration operations for each interval. If the specifications of all gradient pulses are provided to Auto-b, an accurate b-matrix can be obtained. Three examples are explored to validate the accuracy of Auto-b and to detect b-value errors when using approximate calculations. (1) In the conventional spin-echo example, Auto-b exhibits high accuracy, as indicated by the maximum relative deviation of 1.68‰ between its calculated b-matrices and those obtained from analytical expressions. (2) Auto-b is applied to investigate the contribution of imaging gradients to the b-matrix in an optimized spin-echo echo planar imaging sequence at submillimeter resolution. Specifically, ignoring the contribution of imaging gradients results in a b-value error of 12.16 s/mm2 at the 0.8 × 0.8 × 0.8 mm3 resolution and 22.47 s/mm2 at the 0.6 × 0.6 × 0.8 mm3 resolution, respectively, when nominal b = 0. (3) Auto-b is also utilized to analyze the influence of approximate calculations in the spatiotemporally encoded sequence. The results showed that neglecting the contribution of phase-encoding blips causes large b-value errors up to 11.02 s/mm2. In addition, the rectangularization of trapezoidal waveforms led to a high relative b-value error of 39.91%. This study validates the high accuracy of Auto-b and underscores the importance of accurate b-value calculations in both submillimeter imaging and spatiotemporally encoded sequences. Attributed to its automation, accuracy, and broad applicability, Auto-b is helpful for developers of diffusion sequences.

Estimation of hike events and temporal parameters with body-attached sensors

The analysis of human gait is of fundamental importance for the monitoring and enhancement of athletes’ performances. The kinematics and kinetics of human gait are mostly investigated with optical motion capture systems and force plates that require specialised laboratories and limit the possible test conditions. On the contrary, body-attached sensor networks provide an opportunity for long-term acquisitions in unsupervised, naturalistic scenarios. In this study, a wearable sensor network consisting of two wireless dataloggers and two instrumented insoles with eight pressure sensors each is used. Custom algorithms for the automatic detection of hike events and the estimation of the related temporal parameters based on sensors data are presented. The proposed algorithms were tested against laboratory measurements performed on an instrumented treadmill and showed relative errors of less than 2.5% in the estimation of stride time, step time and cadence. Higher relative errors were found in the estimation of stance and swing phases. The developed algorithms were also applied in a field study. In this paper data from one subject are considered. The aim of this research work is to provide an effective sensor-based methodology for the evaluation of gait parameters in naturalistic settings.

A multitask cascading convolutional neural network for high-accuracy pointer meter automatic recognition in outdoor environments

Pointer meter automatic recognition (PMAR) in outdoor environments is a challenging task. Due to variable weather and uneven lighting factors, hand-crafted features or shallow learning techniques have low accuracy in meter recognition. In this paper, a multitask cascading convolutional neural network (MC-CNN) is proposed to improve the accuracy of meter recognition in outdoor environments. The proposed MC-CNN uses cascaded CNN, including three stages of meter detection, meter cropping and meter reading. Firstly, the YOLOV4 Network is used for meter detection to quickly determine the meter location from captured images. In order to accurately cluster pointer meter prior boxes in the YOLOV4 Network, an improved K-means algorithm is presented to further enhance the detection accuracy. Then, the detected meter images are cropped out of the captured images to remove redundant backgrounds. Finally, a meter-reading network based on an adaptive attention residual module (AARM) is proposed for reading meters from cropped images. The proposed AARM not only contains an attention mechanism to focus on essential information and efficiently diminish useless information, but also extracts information features from meter images adaptively. The experimental results show that the proposed MC-CNN can effectively achieve outdoor meter recognition, with high recognition accuracy and low relative error. The recognition accuracy can reach 92.6%. The average relative error is 2.5655%, which is about 3% less than the error in other methods. What is more, the proposed approach can obtain rich information about the type, limits, units and readings of the pointer meter and can be used when multiple pointer meters exist in one captured image simultaneously. Additionally, the proposed approach can significantly improve the accuracy of the recognized readings, and is also robust to natural environments.

An exact version of Life Table Response Experiment analysis, and the R package exactLTRE

Abstract Matrix population models are frequently built and used by ecologists to analyse demography and elucidate the processes driving population growth or decline. Life Table Response Experiments (LTREs) are comparative analyses that decompose the realized difference or variance in population growth rate () into contributions from the differences or variances in the vital rates (i.e. the matrix elements). Since their introduction, LTREs have been based on approximations and have not included biologically relevant interaction terms. We used the functional analysis of variance framework to derive an exact LTRE method, which calculates the exact response of to the difference or variance in a given vital rate, for all interactions among vital rates—including higher‐order interactions neglected by the classical methods. We used the publicly available COMADRE and COMPADRE databases to perform a meta‐analysis comparing the results of exact and classical LTRE methods. We analysed 186 and 1487 LTREs for animal and plant matrix population models, respectively. We found that the classical methods often had small errors, but that very high errors were possible. Overall error was related to the difference or variance in the matrices being analysed, consistent with the Taylor series basis of the classical method. Neglected interaction terms accounted for most of the errors in fixed design LTRE, highlighting the importance of two‐way interaction terms. For random design LTRE, errors in the contribution terms present in both classical and exact methods were comparable to errors due to neglected interaction terms. In most examples we analysed, evaluating exact contributions up to three‐way interaction terms was sufficient for interpreting 90% or more of the difference or variance in . Relative error, previously used to evaluate the accuracy of classical LTREs, is not a reliable metric of how closely the classical and exact methods agree. Error compensation between estimated contribution terms and neglected contribution terms can lead to low relative error despite faulty biological interpretation. Trade‐offs or negative covariances among matrix elements can lead to high relative error despite accurate biological interpretation. Exact LTRE provides reliable and accurate biological interpretation, and the R package exactLTRE makes the exact method accessible to ecologists.

Effect of seepage condition in geological stratification on thermal response test analysis of borehole heat exchanger

Effect of seepage condition in geological stratification on thermal response test analysis of borehole heat exchanger

An Analytical Solution to the One-Dimensional Unsteady Temperature Field near the Newtonian Cooling Boundary

One-dimensional heat-conduction models in a semi-infinite domain, although forced convection obeys Newton’s law of cooling, are challenging to solve using standard integral transformation methods when the boundary condition φ(t) is an exponential decay function. In this study, a general theoretical solution was established using Fourier transform, but φ(t) was not directly present in the transformation processes, and φ(t) was substituted into the general theoretical solution to obtain the corresponding analytical solution. Additionally, the specific solutions and corresponding mathematical meanings were discussed. Moreover, numerical verification and sensitivity analysis were applied to the proposed model. The results showed that T(x,t) was directly proportional to the thermal diffusivity (a) and was inversely proportional to calculation distance (x) and the coefficient of cooling ratio (λ). The analytical solution was more sensitive to the thermal diffusivity than other factors, and the highest relative error between numerical and analytical solutions was roughly 4% under the condition of 2a and λ. Furthermore, T(x,t) grew nonlinearly as the material’s thermal diffusivity or cooling ratio coefficient changed. Finally, the analytical solution was applied for parameter calculation and verification in a case study, providing the reference basis for numerical calculation under specific complex boundaries, especially for the study of related problems in the fields of fluid dynamics and peridynamics with the heat-conduction equation.

Hot workability and microstructure evolution of wire arc additive manufactured 300M steel

Wire arc additive manufacturing (WAAM) technology offers a material-saving and efficient method of manufacturing the 300M steel components, but the obtained microstructure and properties hardly reach the level of wrought materials. By combining WAAM technology with forging process, a novel forming process is proposed. The preferred shape pre-forgings are easily prepared, and the number of forging steps significantly decreases. The WAAMed as-cast microstructure also can be transformed into the wrought state by the proposed forming process. The aim of this study is to investigate the hot deformation behavior and microstructure evolution of WAAMed 300M steel under various deformation temperatures and strain rates. The results show that the WAAMed 300M steel exhibits a higher plastic deformation resistance than the wrought 300M steel. With the increase of deformation temperature and strain rate, the difference of flow stress between the WAAMed and wrought 300M steel decreases. The hot deformation activation energy of WAAMed 300M steel is calculated as 374.1 kJ/mol, which is much higher than that of wrought 300M steel (332.3 kJ/mol). The possible processing windows obtained from the hot deformation activation energy maps are 1040-1120 °C/0.01-10 s−1 for the WAAMed 300M steel and 1010-1130 °C/ 0.1-10 s−1 for the wrought 300M steel. The constitutive model considering strain compensation is established to describe the hot deformation behaviors of WAAMed and wrought 300M steel. The high correlation coefficient and low average absolute relative error confirm the suitability of the developed constitutive models for predicting the flow stresses of WAAMed and wrought 300M steel.

Optimization design of two-stage amplification micro-drive system without additional motion based on particle swarm optimization algorithm

With the increasing requirements of precision mechanical systems in electronic packaging, ultra-precision machining, biomedicine and other high-tech fields, it is necessary to study a precision two-stage amplification micro-drive system that can safely provide high precision and a large amplification ratio. In view of the disadvantages of the current two-stage amplification and micro-drive system, such as poor security, low motion accuracy and limited amplification ratio, an optimization design of a precise symmetrical two-stage amplification micro-drive system was completed in this study, and its related performance was studied. Based on the guiding principle of the flexure hinge, a two-stage amplification micro-drive mechanism with no parasitic motion or non-motion direction force was designed. In addition, the structure optimization design of the mechanism was completed using the particle swarm optimization algorithm, which increased the amplification ratio of the mechanism from 5 to 18 times. A precise symmetrical two-stage amplification system was designed using a piezoelectric ceramic actuator and two-stage amplification micro-drive mechanism as the micro-driver and actuator, respectively. The driving, strength, and motion performances of the system were subsequently studied. The results showed that the driving linearity of the system was high, the strength satisfied the design requirements, the motion amplification ratio was high and the motion accuracy was high (relative error was 5.31%). The research in this study can promote the optimization of micro-drive systems.

Scalable deep learning for watershed model calibration

Watershed models such as the Soil and Water Assessment Tool (SWAT) consist of high-dimensional physical and empirical parameters. These parameters often need to be estimated/calibrated through inverse modeling to produce reliable predictions on hydrological fluxes and states. Existing parameter estimation methods can be time consuming, inefficient, and computationally expensive for high-dimensional problems. In this paper, we present an accurate and robust method to calibrate the SWAT model (i.e., 20 parameters) using scalable deep learning (DL). We developed inverse models based on convolutional neural networks (CNN) to assimilate observed streamflow data and estimate the SWAT model parameters. Scalable hyperparameter tuning is performed using high-performance computing resources to identify the top 50 optimal neural network architectures. We used ensemble SWAT simulations to train, validate, and test the CNN models. We estimated the parameters of the SWAT model using observed streamflow data and assessed the impact of measurement errors on SWAT model calibration. We tested and validated the proposed scalable DL methodology on the American River Watershed, located in the Pacific Northwest-based Yakima River basin. Our results show that the CNN-based calibration is better than two popular parameter estimation methods (i.e., the generalized likelihood uncertainty estimation [GLUE] and the dynamically dimensioned search [DDS], which is a global optimization algorithm). For the set of parameters that are sensitive to the observations, our proposed method yields narrower ranges than the GLUE method but broader ranges than values produced using the DDS method within the sampling range even under high relative observational errors. The SWAT model calibration performance using the CNNs, GLUE, and DDS methods are compared using R2 and a set of efficiency metrics, including Nash-Sutcliffe, logarithmic Nash-Sutcliffe, Kling-Gupta, modified Kling-Gupta, and non-parametric Kling-Gupta scores, computed on the observed and simulated watershed responses. The best CNN-based calibrated set has scores of 0.71, 0.75, 0.85, 0.85, 0.86, and 0.91. The best DDS-based calibrated set has scores of 0.62, 0.69, 0.8, 0.77, 0.79, and 0.82. The best GLUE-based calibrated set has scores of 0.56, 0.58, 0.71, 0.7, 0.71, and 0.8. The scores above show that the CNN-based calibration leads to more accurate low and high streamflow predictions than the GLUE and DDS sets. Our research demonstrates that the proposed method has high potential to improve our current practice in calibrating large-scale integrated hydrologic models.

Determination of salmon freshness by computer vision based on eye color

Determination of salmon freshness by computer vision based on eye color

Determination of Dichloromethane in Blood with Headspace Gas Chromatography and Gas Chromatography-Mass Spectrometry

To establish a method for qualitative determination of dichloromethane (DCM) in blood by gas chromatography-mass spectrometry (GC-MS) and quantitative determination of DCM in blood by headspace gas chromatography (HS-GC), and to provide reliable support for forensic examination and analysis of poisoning or deaths caused by DCM. 0.5 mL blood sample was collected, added into headspace vial with chloroform as the internal standard, and processed by heating at 65 °C and evacuation treatment. The intermediate gas in the headspace vial was analyzed by GC-MS for qualitative validation of the method and by HS-GC for quantitative validation of the method. The method was then applied in forensic case analysis. Qualitative validation of the examination method by GC-MS found that the chromatographic peak and mass spectral characteristic ions were specific in samples added with DCM, and that no interference was observed in the blank negative samples. The limit of detection (LOD) was 5 μg/mL. Quantitative method validation by HS-GC found that the chromatographic peak of DCM was well separated from those of eight other volatile compounds, with the resolution>1.5 in all cases; the lower limit of quantification (LOQ) was 20 μg/mL and good linearity was shown within the range of 20 and 1000 μg/mL, R>0.999; the intra-day test precision and inter-day test precision were good (relative standard deviation, or RSD<15% for both) and test accuracy was high (relative error, or δ<15%). With the method established in the study, DCM was detected successfully in the blood of two fatal cases caused by DCM poisoning, with the blood concentration being 470 μg/mL and 915 μg/mL, respectively. This method is shown to be a rapid, stable and accurate approach to the qualitative and quantitative forensic and toxicological analysis of DCM in blood in DCM poisoning cases or deaths caused by DCM.