Rapid diagnosis of heavy metal pollution in lake sediments based on environmental magnetism and machine learning

doi:10.1016/j.jhazmat.2021.126163

Journal of Hazardous Materials

Volume 416, 15 August 2021, 126163

https://doi.org/10.1016/j.jhazmat.2021.126163 Get rights and content

Highlights

•
Spatiotemporal distribution of heavy metals in lake sediment of Chaohu Lake were analyzed.
•
Magnetic measurements showed that ferrimagnetic minerals are the main magnetic minerals in sediment.
•
Metals were simulated with magnetic parameters and physicochemical indicators as inputs using machine learning approach.
•
Simulation effects of Be, Fe, Pb, Zn, As, Cu and Cr were promising.

Abstract

Environmental magnetism in combination with machine learning can be used to monitor heavy metal pollution in sediments. Magnetic parameters and heavy metal concentrations of sediments from Chaohu Lake (China) were analyzed. The magnetic measurements, high- and low-temperature curves, and hysteresis loops showed the primary magnetic minerals were ferrimagnetic minerals in sediments. For most metals, their concentrations were highest during the wet season and lowest during the medium-water period. Cd, Hg, and Zn were moderately enriched and Cd and Hg posed a considerable ecological risk. A redundancy analysis indicated a relationship between physicochemical indexes and magnetic parameters and heavy metal concentrations. An artificial neural network (ANN) and support vector machine (SVM) were used to construct six models to predict the heavy metal concentrations and ecological risk index. The inclusion of both the physicochemical indexes and magnetic parameters as input factors in the models were significantly ameliorated the simulation accuracy for the majority of heavy metals. The training and test R, for Be, Fe, Pb, Zn, As, Cu, and Cr were > 0.8. The SVM showed better performance and hence it has potential for the efficient and economical long-term tracking and monitoring of heavy metal pollution in lake sediments.

Graphical Abstract

Introduction

The sediments of lakes and streams are major sites of heavy metal deposition and transport and thus comprehensively reflect both the current status and the history of pollution in those water bodies (Zhang et al., 2017a, Zhang et al., 2017b, Kostka and Leśniak, 2020, Niu et al., 2020). Moreover, heavy metals in sediments are a persistent source of biological toxicity with multiple potential impacts, including direct effects on benthic communities and, through their contamination of drinking water and accumulation in aquatic products via the food chain, have indirect, negative effects on human health (Fu et al., 2013, Bing et al., 2013, Huang et al., 2013, Rajeshkumar et al., 2018, Chen et al., 2019). The content of heavy metals in sediments can be monitored by sample collection and large-scale instrument-based analyses in the laboratory, but the required methodologies are time-consuming and expensive (Yang et al., 2009, Bing et al., 2011). The development of an efficient and economical method to study heavy metal pollution in lake sediments would contribute to the improved protection and utilization of water resources, human health, and regional economies.

The difference in the magnetic characteristics of natural materials and secondary substances produced by anthropogenic activities can be exploited to track the impact of the latter on the environment (Wang et al., 2014, Mariyanto et al., 2019, Guda et al., 2020). This strategy has been applied to monitor heavy metal pollution in sediment, soils and leaves, and in the atmosphere. The advantages include less sample consumption, a high sensitivity, simplicity, the use of non-destructive methods, and low cost (Wang et al., 2011, Li et al., 2014, Rachwał et al., 2017, Yang et al., 2019, Wang et al., 2020). The mechanism of correlation between magnetic properties and heavy metals in sediments can be summarized as follows. First of all, exogenous magnetic minerals in lakes are primarily transported by water and the atmosphere, including pollutants produced by human activities, whereas the endogenous magnetic minerals primarily include clastic sediments from lake basins (Jiang et al., 2016). Therefore, magnetic particles have been found to have common sources as heavy metals in sediment (Zhang et al., 2018). Second, Fe/Mn oxides in sediments have the ability to adsorb heavy metals (Fang et al., 2005). Additionally, organic colloids and clay particles can also adsorb magnetic particles and heavy metals at the same time (Fang et al., 2005). Third, the minerals formed by near iron metal elements and the complex of magnetic bacteria both have a certain contribution to the magnetic properties of sediments (Dearing et al., 2001). In recent years, many studies have confirmed the correlation between the magnetic parameters of surface sediments and the total amount of heavy metals (Zhang et al., 2007, Zhang et al., 2011, Guo et al., 2015, Wang et al., 2017, Wang et al., 2020). However, in qualitative or semi-quantitative evaluations of heavy metal pollution, nearly all studies have analyzed only the linear relationship between heavy metals and various magnetic parameters (Wang et al., 2017, Pan et al., 2019, Chaparro et al., 2020). The disadvantage to this approach is that, in addition to the relatively poor accuracy of such linear models, they cannot be used in the assessment for water environments and other inherently non-linear systems.

Machine learning, and especially the support vector machine (SVM) and artificial neural network (ANN), is used to model and predict complex, nonlinear mathematical relationships between independent and dependent variables (Zhang et al., 2019, Bhagat et al., 2020, Šimić et al., 2020). ANN roughly mimics biological neuronal processes by discerning the mapping structure from input data to output data in the absence of an explicit mathematical relationship. While it can be flexibly used to solve relatively complex nonlinear problems (Kakaei Lafdani et al., 2013), it suffers from overfitting, a lack of network optimization, poor generality, iteration failure, and inconsistent results. In contrast, the SVM avoids overfitting and the curse of dimensionality (Li et al., 2017, Parveen et al., 2017). Machine learning has been widely adopted in environmental research, including in the prediction and simulation concentrations of metal elements in atmospheric particles, water, and soil (Liu et al., 2016, Leng et al., 2017, Leng et al., 2018, Jia et al., 2019, Lu et al., 2019). However, due to the influence of complex factors, such as the sources and physicochemical states of heavy metals in sediments, the application of machine learning methods to model heavy metals in lake sediments remains challenging.

In this study, the magnetic parameters, temporal and spatial characteristics, and ecological risk of heavy metals in surface sediment samples collected from Chaohu Lake (China) during different hydrological periods were analyzed. Based on the observed correlation between magnetic parameters and heavy metals, SVM and back-propagation (BP)-ANN models were established in which those magnetic parameters as well as the physicochemical indexes of the lake sediments served as input. The aim of this study was to validate the use of magnetic measurements using a simple, fast, and inexpensive approach for the rapid monitoring of heavy metal concentrations in lake sediments.

Section snippets

Study area

Chaohu Lake (N 31°43′28″–31°25′28″, E 117°16′54″–117°51′46″) is the fifth largest freshwater lake located along the middle reaches of the Yangtze River. It provides drinking water for more than three million people living in the vicinity of the lake. Chaohu Lake has an area of 780 km², an average depth of 2.7 m, and a water storage capacity of 20.7 × 10⁸ m³. The western half of the lake (W) is adjacent to Hefei City, and the eastern half to Chaohu City. The Nanfei, Pai, Baishitian, and Hangbu

Magnetic properties

In determinations of environmental magnetism, χ can reflect the concentration of ferromagnetic minerals and is influenced by paramagnetic and diamagnetic minerals. The SIRM can primarily be attributed to ferrimagnetic minerals and incomplete antiferromagnetic minerals. The HIRM is typically employed to estimate the concentration of high-coercivity minerals. χ_ARM is extremely sensitive to stable single-domain (SSD) and fine-grained pseudo-single domain (PSD) ferrimagnetic minerals. The ratios χ

Conclusion

This study investigated the heavy metal content and magnetic properties of the surface sediments of Chaohu Lake across the hydrological cycle. The concentrations of heavy metals and the ecological risk index in the surface sediments were then simulated using the BP-ANN and SVM models in which physicochemical indexes and magnetic parameters served as input.

Based on their average concentrations the metals were ranked as follows: Fe > Mn > Zn > Cr > Ni > Pb > Cu > Co > As > Be > Sb > Tl > Cd > Hg.

CRediT authorship contribution statement

Xiaolong Li: Performed most of the experiments, analyzed the data, and prepared the final manuscript. Jinxiang Yang: Prepared the final manuscript. Yang Yang: Performed some of the experiments. Yifan Fan: Provided the materials and technical assistance. Xin Qian and Huiming Li: Developed the ideas, designed, and supervised all the experiments.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (grant no. 42077430 and 41771533), the Natural Science Foundation of Jiangsu Province, China (grant no. BK20200716) and the Open Fund of State Key Laboratory of Pollution Control and Resources Reuse (grant no. PCRRF19026).

References (61)

S.K. Bhagat et al.
Development of artificial intelligence for modeling wastewater heavy metal removal: state of the art, application assessment and possible future research
J. Clean. Prod.
(2020)
H. Bing et al.
Historical trends of heavy metal contamination and their sources in lacustrine sediment from Xijiu Lake, Taihu Lake catchment, China
J. Environ. Sci.
(2011)
H. Bing et al.
Assessment of heavy metal enrichment and its human impact in lacustrine sediments from four lakes in the mid-low reaches of the Yangtze River, China
J. Environ. Sci.
(2013)
V. Chandramouli et al.
Backfilling missing microbial concentrations in a riverine database using artificial neural networks
Water Res.
(2007)
R. Chen et al.
Characterization and source apportionment of heavy metals in the sediments of Lake Tai (China) and its surrounding soils
Sci. Total Environ.
(2019)
V. Cherkassky et al.
Practical selection of SVM parameters and noise estimation for SVM regression
Neural Netw.
(2004)
J. Fu et al.
Risk and toxicity assessments of heavy metals in sediments and fishes from the Yangtze River and Taihu Lake
China. Chemosphere
(2013)
J. García-Alba et al.
Artificial neural networks as emulators of process-based models to analyse bathing water quality in estuaries
Water Res.
(2019)
W. Guo et al.
Historical record of human impact in a lake of northern China: magnetic susceptibility, nutrients, heavy metals and OCPs
Ecol. Indic.
(2015)
J. Hu et al.
Influence of different land use types on hydrochemistry and heavy metals in surface water in the lakeshore zone of the Caohai wetland, China
Environ. Pollut.
(2020)

P. Wang et al.

A novel hybrid forecasting model for PM₁₀ and SO₂ daily concentrations

Sci. Total Environ.

(2015)

S. Wang et al.

Environmental magnetic parameter characteristics as indicators of heavy metal pollution in the surface sediments off the Zhoushan Islands in the East China Sea

Mar. Pollut. Bull.

(2020)

Cited by (32)

Interannual variation and machine learning simulation of organophosphate esters in Taihu Lake
2024, Journal of Hazardous Materials
Organophosphate esters (OPEs) are widespread in water bodies and have attracted public attention due to their hazards. This study investigated the presence of OPEs in surface water of Taihu Lake from 2012 and 2021–2022. The OPEs concentration was compared ten years ago and ten years later. Water and meteorological parameters were ranked using the random forest (RF) model, and OPEs concentration in lakes was simulated using selected parameters as inputs. The concentration of Σ₇OPEs was higher ten years ago compared to ten years later. There was no significant seasonal difference in Σ₇OPEs from 2021–2022, while the concentration of Σ₇OPEs in 2012 was lower in summer than in other seasons. The spatial distribution of the two interannual Σ₇OPEs exhibited a decreasing trend from the northwest region. The results of RF importance ranking and redundancy analysis showed that NH₃-N, TN, TP, water temperature and relative humidity were the most influential factors affecting OPEs concentrations. RF models performed better for TnBP, as indicated by training R and test R values are excellent and relatively low errors. Our results demonstrated that machine learning models were useful in facilitating efficient monitoring and assessment of OPEs contamination in lakes.
Estimating the temporal and spatial distribution and threats of bisphenol A in temperate lakes using machine learning models
2024, Ecotoxicology and Environmental Safety
Bisphenol A (BPA) is easily enriched in many human-disturbed watersheds, particularly lakes with poor water mobility, which is posing a threat to aquatic biota. While previous studies have focused on the concentration of BPA in water and its toxicity to aquatic organisms, a small amount of measured data is not enough to reveal the temporal and spatial distribution and threats of BPA, and estimate the ecological risk in watersheds. Therefore, we collected 164 measured BPA data points from Taihu Lake to develop machine learning models using random forest (RF), support vector machine (SVM) and least square regression (LSR) and created month-by-month watershed prediction maps in temperate lakes to estimate the spatiotemporal distribution and threats of BPA. Due to RF's superior robustness to noisy data, the RF model exhibits the best performance among the three algorithms. The RF model showed acceptable predictive performance on the modeling dataset (coefficients of determination and root-mean-square error for the training set were 0.927 and 17.499, respectively, and 0.607, 39.645 for the validation set, respectively). The maps indicated that areas susceptible to anthropogenic activities were more severely polluted by BPA, and rainy climate may favor the migration of BPA to aquatic ecosystems. The model was also applied to predict 42 data points of BPA collected from Dianchi Lake, and the results showed that most predicted data were within a factor of 10 of the measured data, but the prediction accuracy of the model has declined. The ecological risks in the two lakes were evaluated and attention should be paid to the regions with higher risks. Our study provided a novel idea for comprehensive monitoring of an unconventional trace pollutant with endocrine disrupting effects in aquatic ecosystems and analyzing their spatiotemporal distribution, which will contribute to the scientific assessment of the ecological risk of BPA.
Heavy metals and ecological alterations resulting from wastewater discharge in Inner Puno Bay, Lake Titicaca
2023, Environmental Nanotechnology, Monitoring and Management
Wastewater continuously affects lacustrine ecosystems, such is the case of Lake Titicaca, the highest navigable lake in the world. To assess this impact, the concentration of heavy metals in water and sediment samples, physicochemical parameters, and the diversity of phytoplankton were evaluated in locations near the wastewater effluent. With the data on heavy metals in sediments, the Individual Potential Risk Factor Index (E_rⁱ) was assessed and the analysis of Potential Ecological Risk was conducted (PERI). The results revealed that pH, BOD₅, NAMO and Phosphorus exceeded the limits established by Peruvian law. Additionally, we identified 14 metal elements in surface water samples; with a positive correlation between these metals and the sampling point (4), as well as total suspended solids (TSS). In the sediment samples, 35 different metal elements were detected. Specifically, point 4 showed the highest metal contamination levels, while point 9 exhibited the lowest levels. The E_rⁱ in sediments demonstrated that the concentration of cadmium (57.6) represents a moderate risk, while zinc represents the lowest risk (0.0); regarding the PERI, the sampling point closest to the discharge point (4) showed a moderate risk of contamination (113.48), and the point 9 represents the lowest value (14.68); reflecting that sewage affects water quality, negatively impacting phytoplankton diversity. Additionally, the As, Cu, Hg and Pb exceeded the limits set by regulations. Regarding phytoplankton diversity, generally low diversity values were recorded compared to other areas of the lake, with the species Oocystis sp. and Aulacoseira sp prevailing in most of the sampling points. Therefore, urgent measures are needed to remedy the situation in the Inner Puno Bay (IPB) of Lake Titicaca.
Static and dynamic source identification of trace elements in river and soil environments under anthropogenic activities in the Haraz plain, Northern Iran
2023, Science of the Total Environment
Unsustainable human activities have disrupted the natural cycle of trace elements, causing the accumulation of chemical pollutants and making it challenging to determine their sources due to interwoven natural and human-induced processes. A novel approach was introduced for identifying the sources and for quantifying the contribution of trace elements discharge from rivers to soils. We integrated fingerprinting techniques, soil and sediment geochemical data, geographically weighted regression model (GWR) and soil quality indices. The FingerPro package and the state-of-the-art tracer selection techniques including the conservative index (CI) and consensus ranking (CR) were used to quantify the relative contribution of different upland sub-watersheds in trace element discharge soil. Our analysis revealed that off-site sources (upland watersheds) and in-site sources (land use) both play an important role in transferring trace elements to the Haraz plain (northern Iran). The unmixing model's results suggest that the Haraz sub-watersheds exhibit a higher contribution to trace elements transfer in the Haraz plain, and therefore, require greater attention in terms of implementing soil and water conservation strategies. However, it is noteworthy that the Babolroud (adjacent to Haraz) exhibited a better performance of the model. A spatial correlation between certain heavy metals, such as As and Cu, and rice cultivation existed. Additionally, we found a significant spatial correlation between Pb and residential areas, particularly in the Amol region. Our result highlights the importance of using advanced spatial statistical techniques, such as GWR, to identify subtle but critical associations between environmental variables and sources of pollution. The methodology used comprehensively identifies dynamic trace element sourcing at the watershed scale, allowing for pollutant source identification and practical strategies for soil and water quality control. Tracer selection techniques (CI and CR) based on conservatives and consensus improve unmixing model accuracy and flexibility for precise fingerprinting.
Effect of land use conversion on heavy metals and magnetic minerals on water reservoir riparian soils
2023, Chemosphere
Spatial hydrological alterations seem to influence the variability of soil-bound magnetic particles and heavy metals in water reservoir riparian wetlands (RW). To date, applying geochemical analysis with magnetic techniques to assess heavy metal pollution is rarely practiced in RW soils. We studied the magnetic properties and heavy metals, including Cu, Cr, and Zn, of topsoils in RW and the adjacent upland regions (UR, as control) in the Three Gorges Reservoir, China. Potentially elevated low-frequency mass magnetic susceptibility (χ_LF), anhysteretic remanent magnetization susceptibility (χ_ARM), isothermal remanent magnetism, and all selected heavy metals were found in RW. The grain size of the magnetic minerals was coarser in RW than that in UR. The pollution load index (PLI) of the studied samples was 1.18 ± 0.12 and 1.04 ± 0.21 in RW and UR, respectively. PLI and concentrations of Cu, Cr, and Zn were positively correlated with χ_LF, χ_ARM, and isothermal remanent magnetism in RW, whereas no clear linkages were observed between PLI and isothermal remanent magnetism in UR. This finding reveals that hydrological alterations increased the magnetic enhancement and heavy metal enrichment in RW. We find that magnetic proxies of soils could trace the concentration of selected anthropogenic heavy metals and their pollution level in RW.
Improving the efficiency of machine learning in simulating sedimentary heavy metal contamination by coupling preposing feature selection methods
2023, Chemosphere
Sediment cores were collected from Taihu Lake in China. The chronology was determined by radionuclide. Heavy metals and magnetic properties of each core slice were assessed, respectively. The concentrations of most heavy metals in sediments surged at 20 cm from the surface, accompanying the increase in the concentrations of single-domain magnetic particles. This may be resulted from the influence of anthropic activities on the lake's environment after the 1970s. Two feature selection methods, random forest (RF) and maximal information coefficient (MIC), were combined with support vector machine (SVM) model to simulate heavy metals, with the inclusion of selected magnetic and physicochemical parameters. Compared with the modeling results obtained with the full set of parameters, a reasonable simulation performance was obtained with RF and MIC. RF performed better than MIC by increasing the R² of simulation models for Cd, Cr, Cu, Pb, and Sb. For heavy metals with high ecological risks (As, Cd, Cr, Hg, Pb, Sb), the correlation coefficients for observed and predicted data ranged from 0.73 to 0.97 with only 14–27% of the parameters selected by RF as input variables. The RF-RBF-SVM enabled heavy metal predictions based on the magnetic properties of the lake sediments.

View all citing articles on Scopus

View full text

Research PaperRapid diagnosis of heavy metal pollution in lake sediments based on environmental magnetism and machine learning

Highlights

Abstract

Graphical Abstract

Introduction

Section snippets

Study area

Magnetic properties

Conclusion

CRediT authorship contribution statement

Declaration of Competing Interest

Acknowledgments

J. Clean. Prod.

J. Environ. Sci.

J. Environ. Sci.

Water Res.

Sci. Total Environ.

Neural Netw.

China. Chemosphere

Water Res.

Ecol. Indic.

Environ. Pollut.

Chemosphere

Environ. Pollut.

J. Hydrol.

Chemosphere

Chemosphere

Environ. Pollut.

Sci. Total Environ.

Soil Res.

Sci. Total Environ.

Sci. Total Environ.

Sci. Total Environ.

Sci. Total Environ.

Process Saf. Environ.

Geoderma

Chemosphere

Ecotoxicol. Environ. Saf.

Environ. Pollut.

Environ. Pollut.

Sci. Total Environ.

Mar. Pollut. Bull.

Research Paper
Rapid diagnosis of heavy metal pollution in lake sediments based on environmental magnetism and machine learning