Research PaperRapid diagnosis of heavy metal pollution in lake sediments based on environmental magnetism and machine learning
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. χARM 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)
- 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) - 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) - 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) - et al.
Backfilling missing microbial concentrations in a riverine database using artificial neural networks
Water Res.
(2007) - et al.
Characterization and source apportionment of heavy metals in the sediments of Lake Tai (China) and its surrounding soils
Sci. Total Environ.
(2019) - et al.
Practical selection of SVM parameters and noise estimation for SVM regression
Neural Netw.
(2004) - et al.
Risk and toxicity assessments of heavy metals in sediments and fishes from the Yangtze River and Taihu Lake
China. Chemosphere
(2013) - et al.
Artificial neural networks as emulators of process-based models to analyse bathing water quality in estuaries
Water Res.
(2019) - et al.
Historical record of human impact in a lake of northern China: magnetic susceptibility, nutrients, heavy metals and OCPs
Ecol. Indic.
(2015) - 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)
Heavy metal pollution status in surface sediments of Swan Lake lagoon and Rongcheng Bay in the northern Yellow Sea
Chemosphere
A methodological framework for identifying potential sources of soil heavy metal pollution based on machine learning: a case study in the Yangtze Delta, China
Environ. Pollut.
Daily suspended sediment load prediction using artificial neural networks and support vector machines
J. Hydrol.
Spatial and geochemical aspects of heavy metal distribution in lacustrine sediments, using the example of Lake Wigry (Poland)
Chemosphere
Prediction of size-fractionated airborne particle-bound metals using MLR, BP-ANN and SVM analyses
Chemosphere
Leaf magnetic properties as a method for predicting heavy metal concentrations in PM2.5 using support vector machine: a case study in Nanjing, China
Environ. Pollut.
Particulate matter exposure disturbs inflammatory cytokine homeostasis associated with changes in trace metal levels in mouse organs
Sci. Total Environ.
A comprehensive support vector machine-based classification model for soil quality assessment
Soil Res.
Simulating heavy metal concentrations in an aquatic environment using artificial intelligence models and physicochemical indexes
Sci. Total Environ.
Enhance the treatment of low strength wastewater at low temperature with the coexistence system of An AOB and heterotrophic bacteria: performance and bacterial community
Sci. Total Environ.
Heavy metal contents and magnetic properties of surface sediments in volcanic and tropical environment from Brantas River, Jawa Timur Province, Indonesia
Sci. Total Environ.
Meta analysis of heavy metal pollution and sources in surface sediments of Lake Taihu, China
Sci. Total Environ.
Development of SVR-based model and comparative analysis with MLR and ANN models for predicting the sorption capacity of Cr(VI)
Process Saf. Environ.
Application of magnetic susceptibility in assessment of heavy metal contamination of Saxonian soil (Germany) caused by industrial dust deposition
Geoderma
Studies on seasonal pollution of heavy metals in water, sediment, fish and oyster from the Meiliang Bay of Taihu Lake in China
Chemosphere
Artificial intelligence models to predict acute phytotoxicity in petroleum contaminated soils
Ecotoxicol. Environ. Saf.
Applying machine learning methods to better understand, model and estimate mass concentrations of traffic-related pollutants at a typical street canyon
Environ. Pollut.
Detection and differentiation of pollution in urban surface soils using magnetic properties in arid and semi-arid regions of northwestern China
Environ. Pollut.
A novel hybrid forecasting model for PM10 and SO2 daily concentrations
Sci. Total Environ.
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.
Cited by (32)
Interannual variation and machine learning simulation of organophosphate esters in Taihu Lake
2024, Journal of Hazardous MaterialsEstimating the temporal and spatial distribution and threats of bisphenol A in temperate lakes using machine learning models
2024, Ecotoxicology and Environmental SafetyHeavy metals and ecological alterations resulting from wastewater discharge in Inner Puno Bay, Lake Titicaca
2023, Environmental Nanotechnology, Monitoring and Management