A support vector regression model to predict nitrate-nitrogen isotopic composition using hydro-chemical variables

Hazards and influence factors of arsenic in the upper pleistocene aquifer, Hetao region, using machine learning modeling

The Hetao region is one of the regions with the most serious problem of the greatest measured arsenic concentrations in China. The enrichment of arsenic in groundwater may poses a great risk to the health of local residents. A comprehensive understanding of the groundwater quality, spatial distribution characteristics and hazard of the high arsenic in groundwater is indispensable for the sustainable utilization of groundwater resources and resident health. This study selected six environmental factors, climate, human activity, sedimentary environment, hydrogeology, soil, and others, as the independent input variables to the model, compared three machine learning algorithms (support vector machine, extreme gradient boosting, and random forest), and mapped unsafe arsenic to estimate the population that may be exposed to unhealthy conditions in the Hetao region. The results show that nearly half the number of the 605 sampling wells for arsenic exceeded the WHO provisional guide value for drinking water, the water chemistry of groundwater are mainly Na-HCO3-Cl or Na-Mg-HCO3-Cl type water, and the groundwater with excessive arsenic concentration is mainly concentrated in the ancient stream channel influence zone and the Yellow River crevasse splay. The results of factor importance explanation revealed that the sedimentary environment was the key factor affecting the primary high arsenic groundwater concentration, followed by climate and human activities. The random forest algorithm produced the probability distribution of high arsenic groundwater that is consistent with the observed results. The estimated area of groundwater with excessive arsenic reached 38.81 %. An estimated 940,000 people could be exposed to high arsenic in groundwater.

Etemadi regression in chemometrics: Reliability-based procedures for modeling and forecasting

The creation of predictive models with a high degree of generalizability in chemical analysis and process optimization is of paramount importance. Nonetheless, formulating a prediction model based on collected data from chemical measurements that maximize quantitative generalizability remains a challenging task for chemometrics experts. To tackle this challenge, a range of forecasting models with varying characteristics, structures, and capabilities has been developed, utilizing either accuracy-based or reliability-based modeling strategies. While the majority of models follow the accuracy-based approach, a recently proposed reliability-based approach, known as the Etemadi approach, has shown impressive performance across various scientific fields. The Etemadi models were constructed through a reliability-based parameter estimation process in such a manner that maximizes the models' reliability. However, the foundation of modeling procedures for chemometrics purposes is built upon the assumption that high generalizability in inaccessible/test data is achieved through the accuracy-based training procedure in which errors in available historical/training data are minimized. After conducting a thorough review of the current literature, we have found that none of the forecasting models for chemometrics purposes incorporate reliability into their modeling procedures. Given the dynamic and highly sensitive nature of chemistry experiments and processes, implementing a reliable model that controls performance criteria variation is a promising strategy for achieving stable and robust forecasts. To address this research gap, this paper introduces several key innovations, which can be highlighted as follows: (1) Proposing a general design structure based on a new optimal reliability-based parameter estimation process. (2) Introducing a novel risk-based modeling strategy that minimizes the performance variation of models implemented under different conditions in chemical laboratory experiments, to generate a more generalizable model for diverse applications in chemometrics. (3) Specifying the degree of influence that each reliability and accuracy factor has in enhancing the generalizability and uncertainty modeling of chemometric models. Empirical evidence confirms the effectiveness and superior performance of reliability-based models compared to accuracy-based models in 78.95% of the cases across various fields, including Pharmacology, Biochemistry, Agrochemical, Geochemical, Biological, Pollutants, Physicochemical Properties, and Gases Experiment. Furthermore, the study's findings demonstrate that the reliability-based modeling approach outperforms the accuracy-based strategy in terms of MAE, MSE, ARV, and RMSE by an average of 4.697%, 5.646%, 5.646%, and 4.342%, respectively. It is also statistically proven that reliability has a more significant impact on improving the generalizability of chemometric models than accuracy. This emphasizes the importance of including reliability as a crucial factor in chemometrics modeling, a consideration that has been overlooked in traditional modeling processes. Consequently, reliability-based modeling approaches can be regarded as a viable alternative to conventional accuracy-based modeling methods for chemical modeling purposes.

Quantitative prediction of the hydraulic performance of free water surface constructed wetlands by integrating numerical simulation and machine learning

Quantitative prediction of the design parameter-influenced hydraulic performance is significant for optimizing free water surface constructed wetlands (FWS CWs) to reduce point and non-point source pollution and improve land utilization. However, owing to limitations of the test conditions and data scale, a quantitative prediction model of the hydraulic performance under multiple design parameters has not yet been established. In this study, we integrated field test data, mechanism model, statistical regression, and machine learning (ML) to construct such quantitative prediction models. A FWS CW numerical model was established by integrating 13 groups of trace data from field tests. Subsequently, training, test and extension datasets comprising 125 (5^3), 25 (L₂₅(5⁶)) and 16 (L₁₆(4⁴)) data points, respectively, were generated via numerical simulation of multi-level value combination of three quantitative design parameters, namely, water depth, hydraulic loading rate (HLR), and aspect ratio. The short circuit index (φ₁₀), Morrill dispersion index (MDI), hydraulic efficiency (λ) and moment index (MI) were used as representative hydraulic performance indicators. Training set with large samples were analyzed to determine the variation rules of different hydraulic indicators. Based on the control variable method, φ₁₀, λ, and MI grew exponentially with increasing aspect ratio whereas MDI showed a decreasing trend; with increasing water depth, φ₁₀, λ, and MI showed polynomial decreases whereas MDI increased; with increasing HLR, φ₁₀, λ, and MI slowly increased linearly whereas MDI showed the opposite trend. Finally, we constructed models based on multivariate nonlinear regression (MNLR) and ML (random forest (RF), multilayer perceptron (MLP), and support vector regression. The coefficients of determination (R²) of the MNLR and ML models fitting the training and test sets were all greater than 0.9; however, the generalization abilities of different models in the extension set were different. The most robust MLP, MNLR without interaction term, and RF models were recommended as the preferred models to hydraulic performance prediction. The extreme importance of aspect ratio in hydraulic performance was revealed. Thus, gaps in the current understanding of multivariate quantitative prediction of the hydraulic performance of FWS CWs are addressed while providing an avenue for researching FWS CWs in different regions according to local conditions.

Optimization of depth of filler media in horizontal flow constructed wetlands for maximizing removal rate coefficients of targeted pollutant(s)

Varying the depth of HFCW media causes differences in the redox status within the system, and hence the community structure and diversity of bacteria, affecting removal rates of different pollutants. The key functional microorganisms of CWs that remove contaminants belong to the phyla Proteobacteria, Bacteroidetes, Actinobacteria, and Firmicutes. Secondary data of 111 HFCWs (1232 datasets) were analyzed to deduce the relationship between volumetric removal rate coefficients (K_BOD, K_TN, K_TKN, and K_TP) and depth. Equations of depth were derived in terms of rate coefficients using machine learning approach (MLR and SVR) (R² = 0.85, 0.87 respectively). These equations were then used to find the optimum depth for pollutant(s) removal using Grey wolf optimization (GWO). The computed optimum depths were 1.48, 1.71, 1.91, 2.09, and 2.14 m for the removal of BOD, TKN, TN, TP, and combined nutrients, respectively, which were validated through primary data. This study would be helpful for optimal design of HFCWs.

Rapid Determination of Nitrate Nitrogen Isotope in Water Using Fourier Transform Infrared Attenuated Total Reflectance Spectroscopy (FTIR-ATR) Coupled with Deconvolution Algorithm

Nitrate is a prominent pollutant in water bodies around the world. The isotopes in nitrate provide an effective approach to trace the sources and transformations of nitrate in water bodies. However, determination of isotopic composition by conventional analytical techniques is time-consuming, laborious, and expensive, and alternative methods are urgently needed. In this study, the rapid determination of ¹⁵NO₃^- in water bodies using Fourier transform infrared attenuated total reflectance spectroscopy (FTIR-ATR) coupled with a deconvolution algorithm and a partial least squares regression (PLSR) model was explored. The results indicated that the characteristic peaks of ¹⁴NO₃^-/¹⁵NO₃^- mixtures with varied ¹⁴N/¹⁵N ratios were observed, and the proportion of ¹⁵NO₃^- was negatively correlated with the wavenumber of absorption peaks. The PLSR models for nitrate prediction of ¹⁴NO₃^-/¹⁵NO₃^- mixtures with different proportions were established based on deconvoluted spectra, which exhibited good performance with the ratio of prediction to deviation (RPD) values of more than 2.0 and the correlation coefficients (R²) of more than 0.84. Overall, the spectra pretreatment by the deconvolution algorithm dramatically improved the prediction models. Therefore, FTIR-ATR combined with deconvolution and PLSR provided a rapid, simple, and affordable method for determination of ¹⁵NO₃^- content in water bodies, which would facilitate and enhance the study of nitrate sources and water environment quality management.

Comprehensive evaluation of Dendrobium officinale from different geographical origins using near-infrared spectroscopy and chemometrics

Dendrobium officinale, often used as a kind of tea for daily drinks, has drawn increasing attention for its beneficial effects. Quality evaluation of D. officinale is of great significance to ensure its health care value and safeguard consumers' interest. Given that traditional analytical methods for assessing D. officinale quality are generally time-consuming and laborious, this study developed a comprehensive strategy, with the advantages of being rapid and efficient, enabling the quality evaluation of D. officinale from different geographical origins using near-infrared (NIR) spectroscopy and chemometrics. As the quality indicators, polysaccharides, polyphenols, total flavonoids, and total alkaloids were quantified. Three types of wavelength selection methods were used for model optimization and these were synergy interval (SI), genetic algorithm (GA), and competitive adaptive reweighted sampling (CARS). From the qualitative perspective, the geographical origins of D. officinale were differentiated by NIR spectroscopy combined with partial least squares-discriminant analysis (PLS-DA) and support vector classification (SVC). The PLS models constructed based on the wavelengths selected by CARS yielded the best performance for prediction of the contents of quality indicators in D. officinale. The root mean square error (RMSEP) and coefficient of determination (R_p²) in the independent test sets were 12.7768 g kg^-1 and 0.9586, 1.1346 g kg^-1 and 0.9670, 0.3938 g kg^-1 and 0.8803, 0.0825 and 0.7031 and for polysaccharides, polyphenols, total flavonoids, and total alkaloids, respectively. As for the origin identification, the nonlinear SVC was superior to the linear PLS-DA, with the correct recognition rates in calibration and prediction sets up to 100% and 100%, respectively. The overall results demonstrated the potential of NIR spectroscopy and chemometrics in the rapid determination of quality parameters and geographical origin. This study could provide a valuable reference for quality evaluation of D. officinale in a more rapid and comprehensive manner.

A nonlinear autoregressive exogenous (NARX) model to predict nitrate concentration in rivers

Forecasting nitrate concentration in rivers is essential for environmental protection and careful treatment of drinking water. This study shows that nonlinear autoregressive with exogenous inputs neural networks can provide accurate models to predict nitrate plus nitrite concentrations in waterways. The Susquehanna River and the Raccoon River, USA, were chosen as case studies. Water discharge, water temperature, dissolved oxygen, and specific conductance were considered exogenous inputs. The forecasting sensitivity to changes in the exogenous input parameters and time series length was also assessed. For Kreutz Creek at Strickler station (Pennsylvania), the prediction accuracy increased with the number of exogenous input variables, with the best performance achieved considering all the variables (R² = 0.77). The predictions were accurate also for the Raccoon River (Iowa), although only the water discharge was considered exogenous input (South Raccoon River at Redfield-R² = 0.94). Both short- and long-term predictions were satisfactory.

Machine learning exhibited excellent advantages in the performance simulation and prediction of free water surface constructed wetlands

Optimizing the design and operation parameters of free water surface constructed wetlands (FWS CWs) in runoff regulation and wastewater treatment is necessary to improve the comprehensive performance. In this study, nine machine learning (ML) algorithms were successfully developed to optimize the parameter combinations for FWS CWs. The scale effect of surface area on wetland performance was determined based on consistently smaller predictions (-6.2% to -28.9%) of the nine well-established ML algorithms. The models most suitable for FWS CW performance simulation and prediction were random forest and extra trees algorithms because of their high R² values (0.818 in both) with the training set and low mean absolute relative errors (4.7% and 3.8%, respectively) with the test set. Results from feature analysis of the six tree-based algorithms emphasized the importance of water depth and layout of inlet and outlet, and revealed the negligible effect of the aspect ratio. Feature importance and partial dependence analysis enhanced the interpretability of the tree-based algorithms. The proposed ML algorithms enabled the implementation of an extended scenario at a low cost in real time. Therefore, ML algorithms are suitable for expressing the complex and uncertain effects of the design and operation parameters on the performance of FWS CWs. Acquiring datasets consisting of more extensive, uniform, and unbiased parameter combinations is crucial for developing more robust and practical ML algorithms for the optimal design of FWS CWs.

Neuroimaging gradient alterations and epileptogenic prediction in focal cortical dysplasia Ⅲa

INTRODUCTION

Focal cortical dysplasia Type Ⅲa (FCD Ⅲa) is a highly prevalent temporal lobe epilepsy but the seizure outcomes are not satisfactory after epilepsy surgery. Hence, quantitative neuroimaging, epileptogenic alterations, as well as their values in guiding surgery are worth exploring.

METHODS

We examined 69 patients with pathologically verified FCD Ⅲa using multimodal neuroimaging and stereoelectroencephalography (SEEG). Among them, 18 received postoperative imaging which showed the extent of surgical resection and 9 underwent SEEG implantation. We also explored neuroimaging gradient alterations along with the distance to the temporal pole. Subsequently, the machine learning regression model was employed to predict whole-brain epileptogenicity. Lastly, the correlation between neuroimaging or epileptogenicity and surgical cavities was assessed.

RESULTS

FCD Ⅲa displayed neuroimaging gradient alterations on the temporal neocortex, morphology-signal intensity decoupling, low similarity of intra-morphological features and high similarity of intra-signal intensity features. The support vector regression model was successfully applied at the whole-brain level to calculate the continuous epileptogenic value at each vertex (mean-squared error = 13.8 ± 9.8).

CONCLUSION

Our study investigated the neuroimaging gradient alterations and epileptogenicity of FCD Ⅲa, along with their potential values in guiding suitable resection range and in predicting postoperative seizure outcomes. The conclusions from this study may facilitate an accurate presurgical examination of FCD Ⅲa. However, further investigation including a larger cohort is necessary to confirm the results.

Analysis of input set characteristics and variances on k-fold cross validation for a Recurrent Neural Network model on waste disposal rate estimation

The use of machine learning techniques in waste management studies is increasingly popular. Recent literature suggests k-fold cross validation may reduce input dataset partition uncertainties and minimize overfitting issues. The objectives are to quantify the benefits of k-fold cross validation for municipal waste disposal prediction and to identify the relationship of testing dataset variance on predictive neural network model performance. It is hypothesized that the dataset characteristics and variances may dictate the necessity of k-fold cross validation on neural network waste model construction. Seven RNN-LSTM predictive models were developed using historical landfill waste records and climatic and socio-economic data. The performance of all trials was acceptable in the training and validation stages, with MAPE all less than 10%. In this study, the 7-fold cross validation reduced the bias in selection of testing sets as it helps to reduce MAPE by up to 44.57%, MSE by up to 54.15%, and increased R value by up to 8.33%. Correlation analysis suggests that fewer outliers and less variance of the testing dataset correlated well with lower modeling error. The length of the continuous high waste season and length of total high waste period appear not important to the model performance. The result suggests that k-fold cross validation should be applied to testing datasets with higher variances. The use of MSE as an evaluation index is recommended.

Utilizing artificial neural network to simulate and predict the hydraulic performance of free water surface constructed wetlands

Optimizing the hydraulic performance of free water surface constructed wetlands (FWS CWs) is of great economic and ecological value. However, there is a complex nonlinear relationship between the hydraulic performance and design parameters of FWS CWs. In this study, an artificial neural network (ANN) was applied to simulate and predict the hydraulic performance corresponding to different combinations of design parameters, and orthogonal design L₉ (3⁴) was used to determine the optimal combination of the important hyperparameters of the ANN. Based on the convenient scenario prediction ability of ANN, sensitivity analysis of different design parameters was carried out by the control variate method and full factor experiment. The results showed that the combination of 3 hidden layers, 15 neural nodes in each hidden layer, 0.001 learning rate, and 8 batch sizes was optimal for the established ANN model, achieving a coefficient of determination of 0.828 in the validation set and a satisfactory prediction effect in the test set. The narrow feature distribution interval in the training set restricted the generalization ability of the ANN model to some extent. Of the four continuous design parameters, the water depth and aspect ratio had an important influence on the effective volume ratio. The layout of inlet and outlet was the most influential design parameter, as confirmed by the full factor experiment of five factors and four levels. The established ANN allowed real-time implementation in an extended scenario at a low cost. This study suggests that the ANN can simultaneously project complex and uncertain effects of several design parameters on wetland performance. In future research, acquiring further comprehensive, impartial, and unbiased experimental datasets is necessary to establish a more robust and generalizing ANN model that can guide the optimal design of FWS CWs.

Inversion of Soil Salinity Using Multisource Remote Sensing Data and Particle Swarm Machine Learning Models in Keriya Oasis, Northwestern China

Soil salinization is a global problem that damages soil ecology and affects agricultural development. Timely management and monitoring of soil salinity are essential to achieve the most sustainable development goals in arid and semi-arid regions. It has been demonstrated that Polarimetric Synthetic Aperture Radar (PolSAR) data have a high sensitivity to the soil dielectric constant and soil surface roughness, thus having great potential for the detection of soil salinity. However, studies combining PALSAR-2 data and Landsat 8 data to invert soil salinity information are less common. The particle swarm optimization (PSO) algorithm is characterized by simple operation, fast computation, and good adaptability, but there are relatively few studies applying it to soil salinity as well. This paper takes the Keriya Oasis as an example, proposing the PSO-SVR and PSO-BPNN models by combining PSO with support vector machine regression (SVR) and back-propagation neural network (BPNN) models. Then, PALSAR-2 data, Landsat 8 data, evapotranspiration data, groundwater burial depth data, and DEM data were combined to conduct the inversion study of soil salinity in the study area. The results showed that the introduction of PSO generated a satisfactory estimating performance. The SVR model accuracy (R2) improved by 0.07 (PALSAR-2 data), 0.20 (Landsat 8 data), and 0.19 (PALSAR + Landsat data); the BP model accuracy (R2) improved by 0.03 (PALSAR-2 data), 0.24 (Landsat 8 data), and 0.12 (PALSAR + Landsat data), and then combined with the model inversion plots, we found that PALSAR + Landsat data combined with the PSO-SVR model could achieve better inversion results. The fine texture information of PALSAR-2 data can be used to better invert the soil salinity in the study area by combining it with the rich spectral information of Landsat 8 data. This study complements the research ideas and methods for soil salinization using multi-source remote sensing data to provide scientific support for salinity monitoring in the study area.

Hybrid support vector regression and crow search algorithm for modeling and multiobjective optimization of microalgae-based wastewater treatment

Microalgae-based wastewater treatment (and biomass production) is an environmentally benign and energetically efficient technique as compared to traditional practices. The present study is focused on optimization of the major treatment variables such as temperature, light-dark cycle (LD), and nitrogen (N)-to-phosphate (P) ratio (N/P) for the elimination of N and P from tertiary municipal wastewater utilizing Chlorella kessleri microalgae species. In this regard, a hybrid support vector regression (SVR) technique integrated with the crow search algorithm has been applied as a novel modeling/optimization tool. The SVR models were formulated using the experimental data, which were furnished according to the response surface methodology with Box-Behnken Design. Various statistical indicators, including mean absolute percentage error, Taylor diagram, and fractional bias, confirmed the superior performance of SVR models as compared to the response surface methodology (RSM) and generalized linear model (GLM). Finally, the best SVR model was hybridized with the crow search algorithm for single/multi-objective optimizations to acquire the global optimal treatment conditions for maximum N and P removal efficiencies. The best-operating conditions were found to be 29.3°C, 24/0 h/h of LD, and 6:1 of N/P, with N and P elimination efficiencies of 99.97 and 93.48%, respectively. The optimized values were further confirmed by new experimental data.