Machine Learning Applications for Chemical Fingerprinting and Environmental Source Tracking Using Non-target Chemical Data

Experimental insights and modeling innovations: Deciphering Fe(VI) oxidation in imidazole ionic liquids through QSAR and RFR

In this investigation, we conducted a detailed analysis of the oxidation of 16 imidazole ionic liquid variants by Fe(VI) under uniform experimental setups, thereby securing a dataset of second-order reaction rate constants (kobs). This methodology ensures superior data consistency and comparability over traditional methods that amalgamate disparate data from varied studies. Utilizing 16 chemical structural parameters obtained via Density Functional Theory (DFT) as descriptors, we developed a Quantitative Structure Activity Relationship (QSAR) model. Through rigorous correlation analysis, Principal Component Analysis (PCA), Multiple Linear Regression (MLR), and Applicability Domain (AD) evaluation, we identified a pronounced negative correlation between the molecular orbital gap energy (Egap) and kobs. MLR analysis further underscored Egap as a pivotal predictive variable, with its lower values indicating heightened oxidative reactivity towards Fe(VI) in the ionic liquids, leading the QSAR model to achieve a predictive accuracy of 0.95. Furthermore, we integrated an advanced machine learning approach - Random Forest Regression (RFR), which adeptly highlighted the critical factors influencing the oxidation efficiency of imidazole ionic liquids by Fe(VI) through elaborate decision trees, feature importance assessment, Recursive Feature Elimination (RFE), and cross-validation strategies. The RFR model demonstrated a remarkable predictive performance of 0.98. Both QSAR and RFR models pinpointed Egap as a key descriptor significantly affecting oxidation efficiency, with the RFR model presenting lower root mean square errors, establishing it as a more reliable predictive tool. The application of the RFR model in this study significantly improved the model's stability and the intuitive display of key influencing factors, introducing promising advanced analytical tools to the field of environmental chemistry.

Time series (2003-15) analysis of selected physicochemical parameters in Indian Ocean: Cumulative impacts prediction on coral bleaching using machine learning

Coral bleaching is an important ecological threat worldwide, as the coral ecosystem supports a rich marine biodiversity to survive. Sea surface temperature was considered a major culprit; however, later it was observed that other water parameters like pH, tCO2, fCO2, salinity, dissolved oxygen, etc. also play a significant role in bleaching. In the present study, all these parameters of the Indian Ocean area for 15 years (2003-2017) were collected and analysed using machine learning language. The main aim is to see the cumulative impacts of various ocean parameters on coral bleaching. Introducing machine learning in environmental impact assessment studies is a new approach, and the prediction of coral bleaching using simulation of physico-chemical parameters interactions shows 94.4 % accuracy for the prediction of the future bleaching event. This study can be probably the first step in the application of the machine learning language for the prediction of coral bleaching in the field of marine science.

Application of machine learning in the study of development, behavior, nerve, and genotoxicity of zebrafish

Machine learning (ML) as a novel model-based approach has been used in studying aquatic toxicology in the environmental field. Zebrafish, as an ideal model organism in aquatic toxicology research, has been widely used to study the toxic effects of various pollutants. However, toxicity testing on organisms may cause significant harm, consume considerable time and resources, and raise ethical concerns. Therefore, ML is used in related research to reduce animal experiments and assist researchers in conducting toxicological research. Although ML techniques have matured in various fields, research on ML-based aquatic toxicology is still in its infancy due to the lack of comprehensive large-scale toxicity databases for environmental pollutants and model organisms. Therefore, to better understand the recent research progress of ML in studying the development, behavior, nerve, and genotoxicity of zebrafish, this review mainly focuses on using ML modeling to assess and predict the toxic effects of zebrafish exposure to different toxic chemicals. Meanwhile, the opportunities and challenges faced by ML in the field of toxicology were analyzed. Finally, suggestions and perspectives were proposed for the toxicity studies of ML on zebrafish in future applications.

Polypy: A Framework to Interpret Polymer Properties from Mass Spectroscopy Data

Mass spectroscopy (MS) is a robust technique for polymer characterization, and it can provide the chemical fingerprint of a complete sample regarding polymer distribution chains. Nevertheless, polymer chemical properties such as polydispersity (Pd), average molecular mass (Mn), weight average molecular mass (Mw) and others are not determined by MS, as they are commonly characterized by gel permeation chromatography (GPC). In order to calculate polymer properties from MS, a Python script was developed to interpret polymer properties from spectroscopic raw data. Polypy script can be considered a peak detection and area distribution method, and represents the result of combining the MS raw data filtered using Root Mean Square (RMS) calculation with molecular classification based on theoretical molar masses. Polypy filters out areas corresponding to repetitive units. This approach facilitates the identification of the polymer chains and calculates their properties. The script also integrates visualization graphic tools for data analysis. In this work, aryl resin (poly(2,2-bis(4-oxy-(2-(methyloxirane)phenyl)propan) was the study case polymer molecule, and is composed of oligomer chains distributed mainly in the range of dimers to tetramers, in some cases presenting traces of pentamers and hexamers in the distribution profile of the oligomeric chains. Epoxy resin has Mn = 607 Da, Mw = 631 Da, and polydispersity (Pd) of 1.015 (data given by GPC). With Polypy script, calculations resulted in Mn = 584.42 Da, Mw = 649.29 Da, and Pd = 1.11, which are consistent results if compared with GPC characterization. Additional information, such as the percentage of oligomer distribution, was also calculated and for this polymer matrix it was not possible to retrieve it from the GPC method. Polypy is an approach to characterizing major polymer chemical properties using only MS raw spectra, and it can be utilized with any MS raw data for any polymer matrix.

Non-target screening in water analysis: recent trends of data evaluation, quality assurance, and their future perspectives

This trend article provides an overview of recent advancements in Non-Target Screening (NTS) for water quality assessment, focusing on new methods in data evaluation, qualification, quantification, and quality assurance (QA/QC). It highlights the evolution in NTS data processing, where open-source platforms address challenges in result comparability and data complexity. Advanced chemometrics and machine learning (ML) are pivotal for trend identification and correlation analysis, with a growing emphasis on automated workflows and robust classification models. The article also discusses the rigorous QA/QC measures essential in NTS, such as internal standards, batch effect monitoring, and matrix effect assessment. It examines the progress in quantitative NTS (qNTS), noting advancements in ionization efficiency-based quantification and predictive modeling despite challenges in sample variability and analytical standards. Selected studies illustrate NTS's role in water analysis, combining high-resolution mass spectrometry with chromatographic techniques for enhanced chemical exposure assessment. The article addresses chemical identification and prioritization challenges, highlighting the integration of database searches and computational tools for efficiency. Finally, the article outlines the future research needs in NTS, including establishing comprehensive guidelines, improving QA/QC measures, and reporting results. It underscores the potential to integrate multivariate chemometrics, AI/ML tools, and multi-way methods into NTS workflows and combine various data sources to understand ecosystem health and protection comprehensively.

Application of chemical attribution in matching OPNAs-exposed biological samples with exposure sources- based on the impurity profiles via GC × GC-TOFMS analysis

Chemical attribution is a vital tool to attribute chemicals or related materials to their origins in chemical forensics via various chemometric methods. Current progress related to organophosphorus nerve agents (OPNAs) has mainly focused on the attribution of chemical sources and synthetic pathways. It has not yet been applied in matching exposed biological samples to their sources. This work used chemical attribution to explore organic impurity profiles in biological samples exposed to various OPNAs. Chemical attribution was first used to identify the exposure source of biological samples based on the full-scan data via comprehensive two-dimensional gas chromatography-time-of-flight mass spectrometer (GC × GC-TOFMS). Taking peak area as the only variable, it can quickly match exposed samples to their sources by applying unsupervised or supervised models, screen difference compounds via one-way ANOVA or t-tests, and then identify valuable impurities that can distinguish different types of exposed samples. To further obtain the impurity profile only applicable to a certain weapon' samples, the irrelevant components were removed via conventional methods. The findings showed there were 53 impurities that can promote distinguishing six groups of OPNA exposed samples, as well as 42 components that can be used as valuable impurities to distinguish class G and class V samples. These were all unique impurities that appear in a certain weapon' samples. The outcomes can be a reference for tracing the source for OPNA-exposed samples, which was beneficial to the further development in source matching of forensic samples. Moreover, the chemical attribution for impurity profiles in biological samples after weapons exposure may inspire research into the characteristics of impurity profile in biological samples as well as practical applications of chemical attribution for OPNA-exposed samples, that may expand potential biomarkers and break the limits of existing markers in the future.

Identification of organic chemical indicators for tracking pollution sources in groundwater by machine learning from GC-HRMS-based suspect and non-target screening data

In this study, the strong analytical power of gas chromatography coupled to a high resolution mass spectrometry (GC-HRMS) in suspect and non-target screening (SNTS) of organic micropollutants was combined with machine learning tools for proposing a novel and robust systematic environmental forensics workflow, focusing on groundwater contamination. Groundwater samples were collected from four different regions with diverse contamination histories (namely oil [OC], agricultural [AGR], industrial [IND], and landfill [LF]), and a total of 252 organic micropollutants were identified, including pharmaceuticals, personal care products, pesticides, polycyclic aromatic hydrocarbons, plasticizers, phenols, organophosphate flame retardants, transformation products, and others, with detection frequencies ranging from 3 % to 100 %. Amongst the SNTS identified compounds, a total of 51 chemical indicators (i.e., OC: 13, LF: 12, AGR: 19, IND: 7) which included level 1 and 2 SNTS identified chemicals were pinpointed across all sampling regions by integrating a bootstrapped feature selection method involving the bootfs algorithm and a partial least squares discriminant analysis (PLS-DA) model to determine potential prevalent contamination sources. The proposed workflow showed good predictive ability (Q2) of 0.897, and the suggested contamination sources were gasoline, diesel, and/or other light petroleum products for the OC region, anthropogenic activities for the LF region, agricultural and human activities for the AGR region, and industrial/human activities for the IND region. These results suggest that the proposed workflow can select a subset of the most diagnostic features in the chemical space that can best distinguish a specific contamination source class.

Application of Nontarget High-Resolution Mass Spectrometry Fingerprints for Qualitative and Quantitative Source Apportionment: A Real Case Study

High-resolution mass spectrometry (HRMS) provides extensive chemical data, facilitating the differentiation and quantification of contaminants of emerging concerns (CECs) in aquatic environments. This study utilizes liquid chromatography-HRMS for source apportionment in Chebei Stream, an urban water stream in Guangzhou, South China. Initially, 254 features were identified as potential CECs by the nontarget screening (NTS) method. We then established 1689, 1317, and 15,759 source-specific HRMS fingerprints for three distinct sources, the mainstream (C3), the tributary (T2), and the rain runoff (R1), qualitatively assessing the contribution from each source downstream. Subsequently, 32, 55, and 3142 quantitative fingerprints were isolated for sites C3, T2, and R1, respectively, employing dilution curve screening for source attribution. The final contribution estimates downstream from sites C3, T2, and R1 span 32-96, 12-23, and 8-23%, respectively. Cumulative contributions from these sources accurately mirrored actual conditions, fluctuating between 103 and 114% across C6 to C8 sites. Yet, with further tributary integration, the overall source contribution dipped to 52%. The findings from this research present a pioneering instance of applying HRMS fingerprints for qualitative and quantitative source tracking in real-world scenarios, which empowers the development of more effective strategies for environmental protection.

Evidence of transboundary movement of chemicals from Mexico to the U.S. in Tijuana River Estuary sediments

The Tijuana River Estuary (TRE) has been a public health hazard and point of contention between the United States and Mexico for decades, with sources of pollution on both sides of the border. The goal of our study is to determine the presence and dynamics of chemical contamination in the TRE. We sampled sediment from four TRE locations in the U.S. during stable dry conditions and immediately after a wet weather period. Organic chemicals were initially screened with non-targeted analysis using gas chromatography high-resolution mass spectrometry (GC/HRMS) that tentatively identified 6978 chemicals in the NIST 20 database. These tentative identifications were filtered using the USEPA CompTox database to guide quantitative targeted analysis at detection limits below 1 ng/g dry weight sediment. Quantitative targeted analysis of 152 organic pollutants and 18 inorganic elements via GC/HRMS revealed generally higher concentrations of contaminants in dry weather sediments compared to wet weather sediments. The highest concentrations of all chemical classes were detected at the site closest to the U.S.-Mexico border, followed by an urban area near Imperial Beach, California, U.S. All sites exhibited a mixture of petrogenic and pyrogenic polycyclic aromatic hydrocarbons (PAHs). Current-use pesticides were dominated by pyrethroid insecticides and the thiocarbamate herbicide s-Ethyl dipropylthiocarbamate (EPTC), while the U.S.-banned organochlorine pesticides were dominated by chlordanes, dieldrin, and dichlorodiphenyltrichloroethane (DDT) and its degradation byproducts. Polychlorinated biphenyl (PCB) concentrations were greatest at the site closest to the U.S.-Mexico border but in the low nanogram-per-gram range. Phthalates were only found at the same site, with relatively high concentrations of bis(2-ethylhexyl) phthalate. This study provides positive identification and quantitative concentrations for organic pollutants in TRE sediments. Our data suggest that there are multiple sources of chemical contamination in the estuary, including possible transboundary movement of pollutants from Mexico.

Target and suspect per- and polyfluoroalkyl substances in fish from an AFFF-impacted waterway

A major source of toxic per- and polyfluoroalkyl substances (PFAS) is aqueous film-forming foams (AFFF) used in firefighting and training at airports and military installations, however, PFAS have many additional sources in consumer products and industrial processes. A field study was conducted on fish tissues from three reaches of the Columbia Slough, located near Portland International Airport, OR, that are affected by AFFF and other PFAS sources. Fishes including largescale sucker (Catostomus macrocheilus), goldfish (Carassius auratus), and largemouth bass (Micropterus salmoides) were collected in 2019 and 2020. Fish blood, liver, and fillet (muscle) were analyzed for target and suspect PFAS by liquid chromatography high resolution mass spectrometry (LC-HRMS). Data were analyzed for patterns by fish species, tissue type, and river reach. Thirty-three out of 50 target PFAS and additional suspect compounds were detected at least once during the study, at concentrations up to 856 ng/g. Seven carboxylic acids (PFOA, PFNA, PFDA, PFUdA, PFDoA, PFTrDA, PFTeDA), three sulfonates (PFHxS, PFOS, PFDS), three electrofluorination-based compounds (FBSA, FHxSA, FOSA), and two fluorotelomer-based compounds (8:2 FTS, 10:2 FTS) were the most frequently detected compounds in all tissue types. The C6 (PFHxS) to C10 (PFDS) homologs were detected with PFOS and FHxSA at concentrations 1-3 orders of magnitude greater than the other PFAS detected. This is the first report of Cl-PFOS, FPeSA, and FHpSA detected in fish tissue. In all fish samples, fillet concentrations of PFAS were the lowest, followed by liver, and blood concentrations of PFAS were the highest. Differences in PFAS concentrations were driven primarily by tissue types and to a lesser extent fish species, but weakly by river reach. The Oregon Health Authority modified an existing fish consumption advisory on the Columbia Slough to recommend no whole-body consumption of most fish to avoid elevated levels of PFOS in fish liver. Measured PFAS concentrations in fish tissues indicate the potential for adverse ecological effects.

Effect-Directed Analysis Combined with Nontarget Screening to Identify Unmonitored Toxic Substances in the Environment

Effect-directed analysis (EDA) combined with nontarget screening (NTS) has established a valuable tool for the identification of unmonitored toxic substances in environmental samples. It consists of three main steps: (1) highly potent fraction identification, (2) toxicant candidate selection, and (3) major toxicant identification. Here, we discuss the methodology, current status, limitations, and future challenges of EDA combined with NTS. This method has been applied successfully to various environmental samples, such as sediments, wastewater treatment plant effluents, and biota. We present several case studies and highlight key results. EDA has undergone significant technological advancements in the past 20 years, with the establishment of its key components: target chemical analysis, bioassays, fractionation, NTS, and data processing. However, it has not been incorporated widely into environmental monitoring programs. We provide suggestions for the application of EDA combined with NTS in environmental monitoring programs and management, with the identification of further research needs.

Vulnerability of wells in unconfined and confined aquifers to modern contamination from flood events

Groundwater is a primary potable water supply for coastal North Carolina (NC), but the increased intensity of extreme rainfall events and floods may exacerbate surface and subsurface processes that contribute anthropogenic chemicals to wells in the major confined aquifers of this region. We evaluated groundwater for organic chemicals of emerging concern (CEC) and the presence of tritium using flooded and not-flooded wells in the NC Department of Environmental Quality well monitoring network across the NC Coastal Plain. Flooded wells experienced standing water around the well casing at least once during the study period. Tritium concentrations, which indicate modern water presence (water recharged after 1953), were significantly greater in groundwater from flooded wells than not-flooded wells. In confined aquifers, modern water was detected at greater depths in flooded wells (206 m) than not-flooded wells (100 m). Suspect-screening high resolution mass spectrometry (HRMS) analysis of 150 groundwater samples yielded a total of 382 unique organic chemicals. Each groundwater sample contained, on average, 19 tentatively identified chemicals from the NIST 20 mass spectral database (M1) and 9 USEPA ToxCast chemicals. The number of tentatively-identified chemicals per sample was not significantly different among aquifers demonstrating the pervasive presence of the detected CECs in unconfined and confined aquifers. The presence of modern water in groundwater from flooded wells coincided with higher detection frequencies of certain organic contaminant classes, particularly pharmaceuticals, food additives, and regulated aromatic hydrocarbons. These results indicate that wells in both unconfined and confined aquifers are susceptible to modern water contamination during flood events; this finding has critical public health implications for coastal communities.

Nontarget Chemical Composition of Surface Waters May Reflect Ecosystem Processes More than Discrete Source Contributions

We investigated environmental, landscape, and microbial factors that could structure the spatiotemporal variability in the nontarget chemical composition of four riverine systems in the Oregon Coast Range, USA. We hypothesized that the nontarget chemical composition in river water would be structured by broad-scale landscape gradients in each watershed. Instead, only a weak relationship existed between the nontarget chemical composition and land cover gradients. Overall, the effects of microbial communities and environmental variables on chemical composition were nearly twice as large as those of the landscape, and much of the influence of environmental variables on the chemical composition was mediated through the microbial community (i.e., environment affects microbes, which affect chemicals). Therefore, we found little evidence to support our hypothesis that chemical spatiotemporal variability was related to broad-scale landscape gradients. Instead, we found qualitative and quantitative evidence to suggest that chemical spatiotemporal variability of these rivers is controlled by changes in microbial and seasonal hydrologic processes. While the contributions of discrete chemical sources are undeniable, water chemistry is undoubtedly impacted by broad-scale continuous sources. Our results suggest that diagnostic chemical signatures can be developed to monitor ecosystem processes, which are otherwise challenging or impossible to study with existing off-the-shelf sensors.

Interpretation and Prediction of the CO2 Sequestration of Steel Slag by Machine Learning

The utilization of steel slag for CO2 sequestration is an effective way to reduce carbon emissions. The reactivity of steel slag in CO2 sequestration depends mainly on material and process parameters. However, there are many puzzles in regard to practical applications due to the different evaluations of process parameters and the lack of investigation of material parameters. In this study, 318 samples were collected to investigate the interactive influence of 12 factors on the carbonation reactivity of steel slag by machine learning with SHapley Additive exPlanations (SHAP). Multilayer perceptron (MLP), random forest, and support vector regression models were built to predict the slurry-phase CO2 sequestration of steel slag. The MLP model performed well in terms of prediction ability and generalization with comprehensive interpretability. The SHAP results showed that the impact of the process parameters was greater than that of the material parameters. Interestingly, the iron ore phase of steel slag was revealed to have a positive effect on steel slag carbonation by SHAP analysis. Combined with previous literature, the carbonation mechanism of steel slag was proposed. Quantitative analysis based on SHAP indicated that steel slag had good carbonation reactivity when the mass fractions of "CaO + MgO", "SiO2 + Al2O3", "Fe2O3", and "MnO" varied from 50-55%, 10-15%, 30-35%, and <5%, respectively.

Advances and applications of machine learning and deep learning in environmental ecology and health

Machine learning (ML) and deep learning (DL) possess excellent advantages in data analysis (e.g., feature extraction, clustering, classification, regression, image recognition and prediction) and risk assessment and management in environmental ecology and health (EEH). Considering the rapid growth and increasing complexity of data in EEH, it is of significance to summarize recent advances and applications of ML and DL in EEH. This review summarized the basic processes and fundamental algorithms of the ML and DL modeling, and indicated the urgent needs of ML and DL in EEH. Recent research hotspots such as environmental ecology and restoration, environmental fate of new pollutants, chemical exposures and risks, chemical hazard identification and control were highlighted. Various applications of ML and DL in EEH demonstrate their versatility and technological revolution, and present some challenges. The perspective of ML and DL in EEH were further outlined to promote the innovative analysis and cultivation of the ML-driven research paradigm.

Target and Suspect Screening Integrated with Machine Learning to Discover Per- and Polyfluoroalkyl Substance Source Fingerprints

This study elucidates per- and polyfluoroalkyl substance (PFAS) fingerprints for specific PFAS source types. Ninety-two samples were collected from aqueous film-forming foam impacted groundwater (AFFF-GW), landfill leachate, biosolids leachate, municipal wastewater treatment plant effluent (WWTP), and wastewater effluent from the pulp and paper and power generation industries. High-resolution mass spectrometry operated with electrospray ionization in negative mode was used to quantify up to 50 target PFASs and screen and semi-quantify up to 2,266 suspect PFASs in each sample. Machine learning classifiers were used to identify PFASs that were diagnostic of each source type. Four C5-C7 perfluoroalkyl acids and one suspect PFAS (trihydrogen-substituted fluoroethernonanoic acid) were diagnostic of AFFF-GW. Two target PFASs (5:3 and 6:2 fluorotelomer carboxylic acids) and two suspect PFASs (4:2 fluorotelomer-thia-acetic acid and N-methylperfluoropropane sulfonamido acetic acid) were diagnostic of landfill leachate. Biosolids leachates were best classified along with landfill leachates and N-methyl and N-ethyl perfluorooctane sulfonamido acetic acid assisted in that classification. WWTP, pulp and paper, and power generation samples contained few target PFASs, but fipronil (a fluorinated insecticide) was diagnostic of WWTP samples. Our results provide PFAS fingerprints for known sources and identify target and suspect PFASs that can be used for source allocation.

Mass-Suite: a novel open-source python package for high-resolution mass spectrometry data analysis

Mass-Suite (MSS) is a Python-based, open-source software package designed to analyze high-resolution mass spectrometry (HRMS)-based non-targeted analysis (NTA) data, particularly for water quality assessment and other environmental applications. MSS provides flexible, user-defined workflows for HRMS data processing and analysis, including both basic functions (e.g., feature extraction, data reduction, feature annotation, data visualization, and statistical analyses) and advanced exploratory data mining and predictive modeling capabilities that are not provided by currently available open-source software (e.g., unsupervised clustering analyses, a machine learning-based source tracking and apportionment tool). As a key advance, most core MSS functions are supported by machine learning algorithms (e.g., clustering algorithms and predictive modeling algorithms) to facilitate function accuracy and/or efficiency. MSS reliability was validated with mixed chemical standards of known composition, with 99.5% feature extraction accuracy and ~ 52% overlap of extracted features relative to other open-source software tools. Example user cases of laboratory data evaluation are provided to illustrate MSS functionalities and demonstrate reliability. MSS expands available HRMS data analysis workflows for water quality evaluation and environmental forensics, and is readily integrated with existing capabilities. As an open-source package, we anticipate further development of improved data analysis capabilities in collaboration with interested users.

Quantification of Diffusive Methane Emissions from a Large Eutrophic Lake with Satellite Imagery

Lakes are major emitters of methane (CH4); however, a longstanding challenge with quantifying the magnitude of emissions remains as a result of large spatial and temporal variability. This study was designed to address the issue using satellite remote sensing with the advantages of spatial coverage and temporal resolution. Using Aqua/MODIS imagery (2003-2020) and in situ measured data (2011-2017) in eutrophic Lake Taihu, we compared the performance of eight machine learning models to predict diffusive CH4 emissions and found that the random forest (RF) model achieved the best fitting accuracy (R2 = 0.65 and mean relative error = 21%). On the basis of input satellite variables (chlorophyll a, water surface temperature, diffuse attenuation coefficient, and photosynthetically active radiation), we assessed how and why they help predict the CH4 emissions with the RF model. Overall, these variables mechanistically controlled the emissions, leading to the model capturing well the variability of diffusive CH4 emissions from the lake. Additionally, we found climate warming and associated algal blooms boosted the long-term increase in the emissions via reconstructing historical (2003-2020) daily time series of CH4 emissions. This study demonstrates the great potential of satellites to map lake CH4 emissions by providing spatiotemporal continuous data, with new and timely insights into accurately understanding the magnitude of aquatic greenhouse gas emissions.

Digital mapping of soil organic carbon density in China using an ensemble model

The soil organic carbon stock (SOCS) is considered as one of the largest carbon reservoirs in terrestrial ecosystems, and small changes in soil can cause significant changes in atmospheric CO2 concentration. Understanding organic carbon accumulation in soils is crucial if China is to meet its dual carbon target. In this study, the soil organic carbon density (SOCD) in China was digitally mapped using an ensemble machine learning (ML) model. First, based on SOCD data obtained at depths of 0-20 cm from 4356 sampling points (15 environmental covariates), we compared the performance of four ML models, namely random forest (RF), extreme gradient boosting (XGBoost), support vector machine (SVM), and artificial neural network (ANN) models, in terms of coefficient of determination (R2), mean absolute error (MAE), and root mean square error (RMSE) values. Then, we ensembled four models using Voting Regressor and the principle of stacking. The results showed that ensemble model (EM) accuracy was high (RMSE = 1.29, R2 = 0.85, MAE = 0.81), so that it could be a good choice for future research. Finally, the EM was used to predict the spatial distribution of SOCD in China, which ranged from 0.63 to 13.79 kg C/m2 (average = 4.09 (±1.90) kg C/m2). The SOC storage amount in surface soil (0-20 cm) was 39.40 Pg C. This study developed a novel, ensemble ML model for SOC prediction, and improved our understanding of the spatial distribution of SOC in China.

Machine learning in the identification, prediction and exploration of environmental toxicology: Challenges and perspectives

Over the past few decades, data-driven machine learning (ML) has distinguished itself from hypothesis-driven studies and has recently received much attention in environmental toxicology. However, the use of ML in environmental toxicology remains in the early stages, with knowledge gaps, technical bottlenecks in data quality, high-dimensional/heterogeneous/small-sample data analysis and model interpretability, and a lack of an in-depth understanding of environmental toxicology. Given the above problems, we review the recent progress in the literature and highlight state-of-the-art toxicological studies using ML (such as learning and predicting toxicity in complicated biosystems and multiple-factor environmental scenarios of long-term and large-scale pollution). Beyond predicting simple biological endpoints by integrating untargeted omics and adverse outcome pathways, ML development should focus on revealing toxicological mechanisms. The integration of data-driven ML with other methods (e.g., omics analysis and adverse outcome pathway frameworks) endows ML with widely promising application in revealing toxicological mechanisms. High-quality databases and interpretable algorithms are urgently needed for toxicology and environmental science. Addressing the core issues and future challenges for ML in this review may narrow the knowledge gap between environmental toxicity and computational science and facilitate the control of environmental risk in the future.