Model Performance Criteria Research Articles

Development and Validation of a Predictive Model for Growth of Salmonella Infantis in Ground Turkey

Most retail samples (25 g) of ground turkey contain no or low levels of Salmonella. However, temperature abuse after retail can lead to spread and growth of Salmonella in the package. In addition, it can lead to levels that pose a significant risk of salmonellosis. This is especially true when the serotype is a top human clinical isolate, like Infantis. Therefore, the current study was undertaken to develop and validate a predictive model for the growth of Salmonella Infantis in ground turkey subjected to temperature abuse. The purpose was to fill a gap for serotype-specific data and models in risk assessments for this pathogen and food combination. Storage trials with a low initial inoculum (0.85 log10) of Salmonella Infantis in commercial ground turkey samples (0.2 g) with native microflora were conducted at 16–40 °C for 0–28 h. Salmonella was enumerated in ground turkey samples using an automated, whole sample enrichment, miniature, most probable number (MPN) assay. The MPN data were fitted to a three-phase linear primary model. Secondary models for primary model parameters were developed and used in the primary model to create a tertiary model that predicted the growth of Salmonella Infantis in ground turkey as a function of time and temperature. Data and tertiary model predictions were evaluated using the test data, model performance, and model validation criteria of the Acceptable Prediction Zones method in the Validation Software Tool. The tertiary model predictions were considered to have acceptable bias and accuracy when the proportion of residuals (observed − predicted) in the partly and fully acceptable prediction zones (pAPZ) was ≥0.7. The overall pAPZ of the tertiary model was 0.866 for dependent data (n = 406) and 0.853 for independent data for interpolation (n = 177). However, there were local prediction problems that limited the validated prediction range to a region from 0 to 8 h at 16–40 °C. Nonetheless, this validation range was sufficient to simulate temperature abuse of ground turkey during meal preparation in the consumers’ home. Thus, the model fills an important data and modeling gap in risk assessments for Salmonella and ground turkey. Additional data are needed to repair and fully validate the model.

Enhancing Brain Tumor Classification in MRI: Leveraging Deep Convolutional Neural Networks for Improved Accuracy

Brain tumors are among the deadliest forms of cancer, and there is a significant death rate in patients. Identifying and classifying brain tumors are critical steps in understanding their functioning. The best way to treat a brain tumor depends on its type, size, and location. In the modern era, Radiologists utilize Brain tumor locations that can be determined using magnetic resonance imaging (MRI). However, manual tests and MRI examinations are time-consuming and require skills. In addition, misdiagnosis of tumors can lead to inappropriate medical therapy, which could reduce their chances of living. As technology advances in Deep Learning (DL), Computer Assisted Diagnosis (CAD) as well as Machine Learning (ML) technique has been developed to aid in the detection of brain tumors, radiologists can now more accurately identify brain tumors. This paper proposes an MRI image classification using a VGG16 model to make a deep convolutional neural network (DCNN) architecture. The proposed model was evaluated with two sets of brain MRI data from Kaggle. Considering both datasets during the training at Google Colab, the proposed method achieved significant performance with a maximum overall accuracy of 96.67% and 97.67%, respectively. The proposed model was reported to have worked well during the training period and been highly accurate. The proposed model's performance criteria go beyond existing techniques.

Modelling height–diameter relationships in living Araucaria (Araucariaceae) trees to reconstruct ancient araucarian conifer height

AbstractTo reconstruct a fossil forest in three dimensions, an accurate estimation of tree height is crucial. However, modelling the height–diameter relationship of ancient trees is difficult, because the trunks of fossil trees are usually fragmentary, making direct height measurements impossible. One practical approach for reconstructing ancient tree height is to use growth models based on the height–diameter relationships of the nearest living relatives of fossil taxa. Here we apply 19 models to describe height–diameter relationships of living Araucaria trees for establishing appropriate models for ancient Araucariaceae trees. Data come from four living populations of Araucaria: A. bidwillii and A. cunninghamii in Queensland, Australia, and A. cunninghamii and A. hunsteinii in New Guinea. According to an AIC‐based model selection, a power function with an exponent of 0.67 (termed here the modified Mosbrugger model) is found to be the most appropriate for each population and for the entire dataset (157 trees), but normalization constants differ across populations. To find the most appropriate models for the genus Araucaria, 100 random samples (each population generating 25 random samples) from the entire dataset are used. Based on 100 curve fitting results on each model and multiple performance criteria, three median models are generated from the medians of their parameter estimates. Of these, the median 2pPower model works best for Araucaria, but the modified Mosbrugger and Curtis models perform nearly as well. In a case study, we revise tree heights of Upper Jurassic araucariaceous logs in Utah, USA, by applying these three models.

Stock Price Forecasting with Deep Learning Techniques

In this study, LSTM (Long-Short Term Memory) and GRU (Gated Recurrent Unit) techniques of deep learning, which are among the latest advanced technologies, were applied in the Google Colab software program for stock price forecasting. The dataset used in the study was obtained from Yahoo Finance and covers the dates between 02/01/2013 and 30/12/2022. Forecast models were created by considering 5 companies belonging to the XELKT (Electricity Market in Borsa Istanbul) index, which is part of BIST (Borsa Istanbul). Subsequently, the success of these forecast models was tested with the calculated model performance criteria, aiming to determine whether the techniques used were successful in stock price forecasting. Additionally, based on the results of MSE (Mean Squared Error) and MAPE (Mean Absolute Percentage Error) among the calculated model performance criteria, the techniques used were compared with each other, aiming to determine which of these techniques provided forecasts with less error. Then, through the analysis conducted on four different days, an attempt was made to identify the day that yielded the most successful forecasts. As a final step, the goal was to find a model with the least error based on techniques, epoch number, and the number of days forecasted, considering both MSE and MAPE for stocks. Since the model performance criteria outputs obtained from these analyses are below 1 for MSE and below 5% for MAPE, it can be concluded that both techniques demonstrate successful stock price forecasting. Consequently, in the comparison between these two techniques, it is observed that the LSTM technique is slightly more successful than the GRU technique.

Modelling Mixed Traffic Time Headway Distribution Induced by Bus Rapid Transit (BRT) in Cape Town, South Africa

Background: The safety and capacity of roadways are immensely influenced by time headways, a major traffic flow characteristic. Time headways are frequently utilized in various aspects of traffic and transportation engineering studies, including capacity analyses, studies of safety, modelling of lane-changing and car-following behaviour, and level of service assessment. Time headway, measured in seconds, refers to the distance between two subsequent passing cars moving over a single spot on the road. This paper reports the statistical modelling of the time headway distribution of a mixed traffic flow scenario caused by BRT via statistical headway distribution models. The paper posits that the introduction of BRT dedicated lane and its adopted design configuration induces anomalous traffic flows on its adjoining lanes with attendant time headway differentials, irrespective of the type of adopted design configuration for the corridor. The objectives of this study were to fit the BRT-induced time headway distribution to probability distribution models and determine the most appropriate distribution model that fits the data derived using the goodness of fit tests. Consequently, this contribution, therefore, fills the gap in research with respect to determining the precise model that fits the time headway distribution of mixed traffic flow scenarios induced by BRT. Methods: The time headway distribution data collected at four designated road segments on a multilane provincial route R27 caused by BRT dedicated lane in Cape Town, South Africa, were fitted to five probability distribution models viz, Lognormal, Inverse Gaussian, Log-logistic, Generalized Extreme Value (GEV), and Burr using MATLAB software. The fitted headway data were with respect to the traffic flows on the adjoining lanes to BRT or traffic flow ‘without BRT’. Precisely, the empirical data were collected over a three-month period using an Automatic Traffic Counter (ATC). Using ModelRisk software, the goodness of fit of the probability models was assessed by the Akaike Information Criterion (AIC), the Schwartz or Bayesian Information Criterion (SIC or BIC), the Hannan Quinn Information Criterion (HQIC), and the loglikelihood (LLH) model performance criteria respectively. The distribution model with the lowest AIC criterion and largest loglikelihood (LLH) values describe the model that best fits the headway data across the four sites. Results: Results showed that the models fitted the headway data well at each site by visual inspection of the attendant Probability Density Function (PDF) and the Probability plots (P-P). However, the Burr distribution provided the overall best fit based on the AIC and LLH values at 95% confidence and 0.05 significance levels across the four sites. At sites 02, 03, and 04, it ranked first with the lowest AIC values of 4025.99, 2595.56, 3815.36, and corresponding largest LLH values of -2008.98, -1293.76, and -1903.66, respectively, while the lognormal distribution performed best at site 01, with AIC value of 4445.14 and LLH value of -2220.57, closely followed by the Burr distribution with AIC and LLH values of 4453.05 and -2222.51, respectively. The P-values, which ranged between 0.65 and 0.80, showed the likelihood of the occurrence of the data sets under the null hypothesis. Hence the null hypothesis was accepted. Conclusion: The study concluded that the introduction of BRT dedicated lanes affects the adjoining lanes’ mixed traffic time headways’ distribution, and the headways are continuously distributed, hence fitting continuous probability distributions.

A validation of WEPP water quality routines in uniform and nonuniform agricultural hillslopes

Current watershed-scale, nonpoint source pollution models do not represent the processes and impacts of agricultural best management practices on water quality with sufficient detail. A Water Erosion Prediction Project-Water Quality (WEPP-WQ) model was recently developed which is capable of simulating nonpoint source pollutant transport in nonuniform hillslope conditions such as those with BMPs. However, WEPP-WQ has not been validated for these conditions, and prior validation work only evaluated calibrated performance rather than uncalibrated performance, with the latter being most relevant to model applications. This study evaluated uncalibrated and calibrated model performance in two plot-scale, artificial rainfall studies. 179 observations were compared to corresponding WEPP-WQ simulations of runoff, sediment yield, and soluble and particulate nutrient forms for both nitrogen and phosphorus. Uncalibrated validation results were mixed for the different field conditions, model configurations, and prediction variables. Nash-Sutcliffe Efficiencies for uncalibrated simulations of uniform conditions were generally greater than 0.6 except for soluble nitrogen predictions which were poor. Simulations of nonuniform conditions were generally ‘unsatisfactory’ except for runoff predictions which were quite good (NSE = 0.78). Performance was improved substantially for almost all endpoints with calibration. Some exceptions to this occurred because the objective function for calibration was based on log-space differences so as to more equally-weight calibration of unsaturated conditions that tend to produce lesser runoff volumes and sediment yields. Calibrated results for both uniform and nonuniform conditions were generally ‘satisfactory’ or ‘good’ according to widely accepted model performance criteria.

Evaluation of the effect of scenarios in the 6th report of IPCC on the prediction groundwater level using the non-linear model of the input-output time series.

Due to the increase in greenhouse gases, water and climate crises, increasing population, and decreasing water resources, accurately predicting the changes in the GWL is essential for the management of water resources. For this purpose, in this research, the MIROCES2L model was used to predict the climatic parameters of Birjand Plain under three scenarios of the sixth climate change report: SSP1-2.6, SSP2-4.5, and SSP5-8.5. The minimum temperature, maximum temperature, and precipitation parameters from these three scenarios were measured using the CMhyd model. The results indicated that the minimum and maximum temperature would generally increase in the future under the influence of climate change, but precipitation has a sinusoidal behavior and has a decreasing trend in the summer and spring seasons and an increasing trend in the winter and autumn seasons. Then, three ANN, NIO, and MLR models were employed to simulate groundwater depletion. The results indicated that the evaluation of the performance criteria of the NIO model is superior to the other two models, and it was chosen as the model for predicting groundwater depletion in the future period under the influence of climate change based on all three mentioned scenarios. The final results of this research indicated that the GWL of Birjand Plain in the future period (2024-2041) under the SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios would respectively decrease to 5.58m, 5.13m, and 5.38. The results of this research indicate that the need for sustainable management to conserve groundwater resources is also very important in the study area.

Investigation of Turkish Students' School Engagement through Random Forest Methods Applied to TIMSS 2019: A Problem of School Psychology

In this study, in line with TIMSS 2019 data, 8th grade students' school engagement in the Turkish education system and various variables in the field of science and mathematics, which are thought to be related to school engagement, were examined with the Random Forest method, a method amongst data mining methods. This research focuses on student bullying, which includes psychological factors at school, such as whether students like or dislike their lessons, their confidence in science, their self-confidence, absenteeism, and the effects of teachers' teaching methods on SE. The sample of the study consisted of 3872 students in the science data set and 3802 students in the mathematics data set, which remained as a result of the lost data deletion and assignment processes from 4077 students who originally participated in the application. The open-source Python infrastructure was used in the analysis of the data. Orange 3.32 data mining program was employed for model setup. The model performance criteria MSE, RMSE, MAE and R2 values obtained as a result of the analysis. In both areas, the variables that contribute to the prediction of students' school engagement were ranked according to their importance levels, starting from the most important, also interpreted and discussed. It was observed that the performance criteria of the established model have values close to zero in the field of science (MSE: 2.775 RMSE: 1.666 MAE: 1.267) and mathematics (MSE: 2.240 RMSE: 1.497 MAE: 1.131). Variables explain school engagement at the rate of 69.6% in science and 75.7% in mathematics. The order of importance of the variables in both areas showed a great similarity. Student bullying was obtained as the most important variable. Prospective studies can also be planned towards collecting more in-depth data deploying qualitative data collection methods under a qualitative research model that also includes the opinions of self, peers, teachers and parents. For the education policies that the TES should produce toward increasing SE, it is weighty to reduce and prevent student bullying in schools with sustainable practices. Exclusively school administrations should focus on student bullying with the help of counselors, diagnose the problems and take measures in and out of school according to the types of bullying.

An ensemble tree-based prediction of Marshall mix design parameters and resilient modulus in stabilized base materials

A flexible pavement with an adequate Marshall mix design for the asphalt mixture surface layer(s) and appropriate subbase/base design offers a proper pavement structure for driving safety and comfortability. The conventional approach of calculating the Marshall mix design properties (e.g., stability and flow) and resilient modulus of base materials typically requires laborious, expensive, and time-consuming laboratory sample preparation and testing, such as the Marshall stability test and Resilient modulus test. Therefore, this study conducts research on the application of tree-based ensemble methods specifically, random forest (RFR) and eXtreme gradient boosting regression (XGBR) to predict the properties (mainly including Marshall stability, flow, VMA, VFA, and unit weigh) of asphalt mix following Marshall mix design and resilient modulus (Mr) of stabilized aggregate base material. To develop the prediction models, a comprehensive database was established using existing literature. The performance of the prediction models was evaluated using statistical parameters, performance index (p), and other suggested model performance criteria from external research. Additionally, the Shapley Additive exPlainations (SHAP) interpretation technique was employed to explain the models. The analysis results demonstrate that both RFR and XGBR methods have the excellent fitting capability (R > 0.9). However, XGBR has superior performance in predicting Marshall mix properties and Mr value. Comparison with previous research studies reveals that the proposed models show superior performance than genetic programming-based Marshall mix design models. The RFR model performs approximately similar to ANN, while XGBR outperforms ANN, and particle swarm optimization-extreme learning machine-based models for Mr prediction. Furthermore, the equation developed based on multi expression programming for Mr value prediction exhibits superior performance compared to previous research study gene expression programming model.

Comparison of GARCH and SVRGARCH models: Example of gold return

Gold has been a precious resource for people on earth from the past to the present. It is used as both a value gain and jewelry, and is the focus of interest for people in terms of receiving attention and protecting its value. Especially recently, it has been the most favorite for investors due to its excess value increase and decrease which is constantly monitored. The study aimed to compare the predictive performance of the gold price return using the Support Vector Regression-GARCH hybrid models combined with the traditional volatility models. It has been examined whether the Support Vector Regression GARCH models would increase foresight performance. The study used data on the daily frequency between 01/01/2010–01/04/2023. Generalized Autoregressive Conditional Variable Variance, Glosten-Jaganthan-Runkle GARCH, Exponential GARCH and hybrid model Support Vector Regression -GARCH are utilized as prediction methods. For all methods, the gold series is divided into two groups as training and test data. The Root Mean Square Error values are compared as a model performance criterion. The RMSE values and graphics outputs have been concluded that the Support Vector Regression-GARCH model based on predicted linear, radial-based and polynomial kernel predicts more effectively than the GARCH models.

Decay rate estimation of respiratory viruses in aerosols and on surfaces under different environmental conditions

Majority of the viral outbreaks are super-spreading events established within 2-10h, dependent on a critical time interval for successful transmission between humans, which is governed by the decay rates of viruses. To evaluate the decay rates of respiratory viruses over a short span, we calculated their decay rate values for various surfaces and aerosols. We applied Bayesian regression and ridge regression and determined the best estimation for respiratory viruses, including severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), severe acute respiratory syndrome coronavirus (SARS-CoV), middle east respiratory syndrome coronavirus (MERS-CoV), influenza viruses, and respiratory syncytial virus (RSV); the decay rate values in aerosols for these viruses were 4.83±5.70, 0.40±0.24, 0.11±0.04, 2.43±5.94, and 1.00±0.50 h-1, respectively. The highest decay rate values for each virus type differed according to the surface type. According to the model performance criteria, the Bayesian regression model was better for SARS-CoV-2 and influenza viruses, whereas ridge regression was better for SARS-CoV and MERS-CoV. A simulation using a better estimation will help us find effective non-pharmaceutical interventions to control virus transmissions.

Leave-one-out cross-validation, penalization, and differential bias of some prediction model performance measures—a simulation study

BackgroundThe performance of models for binary outcomes can be described by measures such as the concordance statistic (c-statistic, area under the curve), the discrimination slope, or the Brier score. At internal validation, data resampling techniques, e.g., cross-validation, are frequently employed to correct for optimism in these model performance criteria. Especially with small samples or rare events, leave-one-out cross-validation is a popular choice.MethodsUsing simulations and a real data example, we compared the effect of different resampling techniques on the estimation of c-statistics, discrimination slopes, and Brier scores for three estimators of logistic regression models, including the maximum likelihood and two maximum penalized likelihood estimators.ResultsOur simulation study confirms earlier studies reporting that leave-one-out cross-validated c-statistics can be strongly biased towards zero. In addition, our study reveals that this bias is even more pronounced for model estimators shrinking estimated probabilities towards the observed event fraction, such as ridge regression. Leave-one-out cross-validation also provided pessimistic estimates of the discrimination slope but nearly unbiased estimates of the Brier score.ConclusionsWe recommend to use leave-pair-out cross-validation, fivefold cross-validation with repetitions, the enhanced or the .632+ bootstrap to estimate c-statistics, and leave-pair-out or fivefold cross-validation to estimate discrimination slopes.

Comparison of neural network, Gaussian regression, support vector machine, long short-term memory, multi-gene genetic programming, and M5 Trees methods for solving civil engineering problems

In this study, it was investigated that how machine learning (ML) methods show performance in different problems having different characteristics. Six ML approaches including Artificial neural networks (ANN), gaussian process regression (GPR), support vector machine regression (SVMR), long short-term memory (LSTM), multi-gene genetic programming (MGGP) and M5 model tree (M5Tree) were utilized to analyze three independent civil engineering problems belonging to construction management, geotechnical engineering, and hydrological engineering sub-disciplines. Mean absolute percentage error (MAPE), root mean square error (RMSE), coefficient of determination (R2), relative root means square error (RRMSE), Nash–Sutcliffe efficiency (NSE), Kling–Gupta efficiency (KGE), and overall index of model performance (OI) criteria were used to evaluate the performances of the models. Besides performance criteria, the relative performances of the six ML models were assessed using Taylor diagram, Violin diagram and One-Tailed Wilcoxon Signed-Rank Test. For each of the problem considered in this study, the effectiveness of the input parameters on the output parameter has been defined using the Relief Method and Correlation Coefficient. The results show that ANN and MGGP models yielded the most successful estimations for three different problems considered. The best prediction was achieved by MGGP model for hydrological engineering problem. For the construction management, geotechnical engineering problems, the best results were obtained using the ANN model. All models were reliable to solve the geotechnical engineering and hydrological engineering problems while LSTM and SVMR models are not reliable to solve the construction management problem. The most and least effective input parameters on output parameter were contract cost (CC) and work definition number (WDN) for the managerial data set. On the other hand, the most and least effective input parameters on the output parameters for the experimental and natural data sets have been obtained as width of the pile (B), rotation degree (R) and minimum temperature (Tmin), streamflow (Q) data, respectively. The number of data and data selection have a significant effect on the homogeneity of the data set and its representativeness of the problem. The error values obtained in test stage are affected from this condition. The equations to calculate the outputs of each of the problem considered were obtained using MGGP and M5Tree models.

Taxonomic bias in freshwater phytoplankton communities and its effect on environmental assessment metrics

ABSTRACT Ecological studies of the highly diverse groups of organisms, such as algae, can be particularly prone to taxonomic opinion bias. There is mounting pressure to recognize and report this bias across studies to facilitate reproducible science and data re-use; however, there are fewer studies which link taxonomic bias to loss of precision in modelling community responses or in bioassessments. We use long-term phytoplankton monitoring data to evaluate effects of taxonomic bias on relevant ecological metrics including algal biovolume, cell counts, diversity, and the performance of a species-based weighted-averaging model for inferring phosphorus concentrations. Despite the non-zero taxonomic bias between paired original and quality assurance samples, several important indicators derived from these data, including model-inferred phosphorus concentrations, total algal biovolume and cell density, and diversity were mostly unaffected. We discuss this as an example of using ecologically relevant model performance criteria for making recommendations about the acceptable level of taxonomic bias.

Modeling Functional Flows in California’s Rivers

Environmental flows are critical to the recovery and conservation of freshwater ecosystems worldwide. However, estimating the flows needed to sustain ecosystem health across large, diverse landscapes is challenging. To advance protections of environmental flows for streams in California, United States, we developed a statewide modeling approach focused on functional components of the natural flow regime. Functional flow components in California streams—fall pulse flows, wet season peak flows and base flows, spring recession flows, and dry season baseflows—support essential physical and ecological processes in riverine ecosystems. These functional flow components can be represented by functional flow metrics (FFMs) and quantified by their magnitude, timing, frequency, duration, and rate-of-change from daily streamflow records. After calculating FFMs at reference-quality streamflow gages in California, we used machine-learning methods to estimate their natural range of values for all stream reaches in the state based on physical watershed characteristics, and climatic factors. We found that the models performed well in predicting FFMs in streams across a diversity of landscape and climate contexts, according to a suite of model performance criteria. Using the predicted FFM values, we established initial estimates of ecological flows that are expected to support critical ecosystem functions and be broadly protective of ecosystem health. Modeling functional flows at large regional scales offers a pathway for increasing the pace and scale of environmental flow protections in California and beyond.

Prediction of Human Development Index with Health Indicators Using Tree-Based Regression Models

Machine learning is a field of artificial intelligence that allows computers to predict and model future events by making inferences from past information with mathematical and statistical operations. In this study, we used tree-based regression models, one of the machine learning methods, to determine and predict the effect of health indicators of 191 countries on the human development index (HDI) between 2014 and 2018 years. When tree-based regression models were compared according to model performance criteria, it was found that the best model was the gradient boosting model with the highest R2 = 0.9962 and the smallest RMSE = 0.0094. With the gradient boosting model, the three most important variables to HDI are; current health expenditure per capita, physicians and nurses, and midwives, respectively. By selecting the ten countries with the highest HDI values and Turkey, HDI values were estimated for 2018-2019 with a gradient boosting model. The countries for which HDI values are best predicted by the gradient boosting method are Netherlands, Sweden, Norway, Iceland, Denmark, Turkey, Ireland, Germany, Australia, and China.

Influence of emission inventory resolution on the modeled spatio-temporal distribution of air pollutants in Buenos Aires, Argentina, using WRF-Chem

The temporal and spatial resolution of the emission inventory included into an air quality model plays a key role in the appropriate representation of air pollution events and background atmospheric chemistry. Here, we use the Weather Research and Forecasting coupled with Chemistry (WRF-Chem v4.0) model to perform high-resolution air quality simulations over the city of Buenos Aires, Argentina, with two different anthropogenic emissions datasets: the High-resolution Emissions Inventory of Argentina (GEAA-AEI) and the Emissions Database for Global Atmospheric Research - Hemispheric Transport of Air Pollution (EDGAR-HTAP). A local optimized configuration considering 3 nested domains with a horizontal grid size of 20 × 20 km, 4 × 4 km, and 1.3 × 1.3 km and the MOZART chemical scheme was used. The model performance for NO2, PM10, PM2.5, and O3 concentrations was validated against measurements from the existing air quality monitoring stations in the Buenos Aires Metropolitan Area (AMBA) during austral fall 2018. Our results show that the daytime concentrations of air pollutants are influenced by the shape and shift of the hourly emissions profile, especially for NO2 where the reduction in nighttime emissions decreased the mean model bias by ∼50%. PM10 and PM2.5 generally satisfied the model performance criteria, but underestimation tended to occur in the GEAA-AEI simulations and overestimation for the EDGAR-HTAP case. Comparison with TROPOMI-derived tropospheric NO2 columns showed a high positive correlation (r > 0.75) and a positive bias. We found large discrepancies between the spatial distribution patterns of the simulations within the innermost high-resolution domain centered on AMBA, mostly in suburban areas where no observations are available. We propose additional monitoring sites to address such differences and determine the size and shape of the main pollutant plume. We conclude that high-resolution air quality modeling is important within underdeveloped or developing South American cities that lack continuous air quality measurements, as it represents a powerful tool in supporting the design of governmental monitoring networks and air pollution mitigation policies.

Development and validation of a neural network model for growth of Salmonella Newport from chicken on cucumber for use in risk assessment

A neural network model was developed for predicting growth of a chicken isolate of Salmonella Newport on cucumber portions as a function of times (0 to 8 hr) and temperatures (16 to 40℃) observed during meal preparation and serving for use in risk assessment. Model development and validation were accomplished using the test data, model performance, and model validation criteria of the Acceptable Prediction Zones (APZ) method in the Validation Software Tool (ValT). The model was considered to provide acceptable predictions when the proportion of residuals in the APZ (pAPZ) was ≥0.70. Data for model development (n = 140) and validation (n = 72) satisfied all criteria of the APZ method in ValT with pAPZ of 0.97 and 0.93, respectively. Thus, the model was successfully validated and can be used with confidence in risk assessment to predict growth of Salmonella Newport from chicken on cucumber during meal preparation and serving. Novelty impact statement The model can be used by risk assessors to help predict variability and uncertainty of consumer exposure to Salmonella from individual lots of chicken produced by a farm-to-table scenario that includes cross-contamination of sliced cucumbers with Salmonella Newport from raw chicken followed by growth of Salmonella Newport on cucumber for times and temperatures observed during meal preparation and serving.

An electric circuit model for a lithium-ion battery cell based on automotive drive cycles measurements

The on-board energy storage system plays a key role in electric vehicles since it directly affects their performance and autonomy. The lithium-ion battery offers satisfactory characteristics that make electric vehicles competitive with conventional ones. This article focuses on modeling and estimating the parameters of the lithium-ion battery cell when used in different electric vehicle drive cycles and styles. The model consists of an equivalent electrical circuit based on a second-order Thevenin model. To identify the parameters of the model, two algorithms were tested: Trust-Region-Reflective and Levenberg-Marquardt. To account for the dynamic behavior of the battery cell in an electric vehicle, this identification is based on measurement data that represents the actual use of the battery in different conditions and driving styles. Finally, the model is validated by comparing simulation results to measurements using the mean square error (MSE) as model performance criteria for the driving cycles (UDDS, LA-92, US06, neural network (NN), and HWFET). The results demonstrate interesting performance mostly for the driving cycles (UDDS and LA-92). This confirms that the model developed is the best solution to be integrated in a battery management system of an electric vehicle.

Derivation of S-Curve from Oscillatory Hydrograph Using Digital Filter

An oscillatory S-curve causes unexpected fluctuations in a unit hydrograph (UH) of desired duration or an instantaneous UH (IUH) that may affect the constraints for hydrologic stability. On the other hand, the Savitzky–Golay smoothing and differentiation filter (SG filter) is a digital filter known to smooth data without distorting the signal tendency. The present study proposes a method based on the SG filter to cope with oscillatory S-curves. Compared to previous conventional methods, the application of the SG filter to an S-curve was shown to drastically reduce the oscillation problems on the UH and IUH. In this method, the SG filter parameters are selected to give the minimum influence on smoothing and differentiation. Based on runoff reproduction results and performance criteria, it appears that the SG filter performed both smoothing and differentiation without the remarkable variation of hydrograph properties such as peak or time-to peak. The IUH, UH, and S-curve were estimated using storm data from two watersheds. The reproduced runoffs showed high levels of model performance criteria. In addition, the analyses of two other watersheds revealed that small watershed areas may experience scale problems. The proposed method is believed to be valuable when error-prone data are involved in analyzing the linear rainfall–runoff relationship.