Risk assessment of groundwater arsenic in Hetao Basin base on ensemble learning optimization

doi:10.13745/j.esf.sf.2023.2.40

Abstract

Abstract:

The shallow groundwater arsenic pollution in Hetao Basin seriously exceeds the standard, and its potential pollution risk poses a serious health threat to local residents. At present, the perception of the risk distribution of high arsenic groundwater is still insufficient on the macroscopic scale. Based on 605 shallow groundwater samples and environmental factors such as sedimentary environment, climate, human activities, soil physical and chemical characteristics, and hydrogeological conditions as data sources, Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Support Vector Machine (SVM) were selected as the base learners, and Linear Discriminant Analysis (LDA) was selected as the meta-learner to construct a Stacking ensemble learning model for high arsenic groundwater. The ensemble learning model was used to predict the risk distribution of high arsenic groundwater and identify the key environmental factors affecting the risk distribution of high arsenic groundwater in the region. The research showed that the groundwater arsenic concentration exceeded the standard rate (>10 μg/L) was 49.59%, mainly concentrated in the paleochannel zone and flood fans of the Yellow River. The Stacking ensemble model had higher reliability than the RF model with the best performance in the single model, and the Area Under the ROC Curve (AUC) and accuracy were increased by 1.1% and 3.2%, respectively. The high-risk area reached 5257 km², accounting for 38.44% of the total area of the study area. The sedimentary environment is the key environmental factor affecting the risk distribution of high arsenic groundwater, contributing up to 25.06% to the accuracy of the model. The results of this study can provide a method and reference for mapping the spatial distribution of high arsenic groundwater pollution and have important implications for the safety of drinking water and human health in the region.

Key words: Stacking ensemble learning, groundwater, high arsenic, risk distribution, Hetao Basin

CLC Number:

FU Yu, CAO Wengeng, ZHANG Chunju, ZHAI Wenhua, REN Yu, NAN Tian, LI Zeyan. Risk assessment of groundwater arsenic in Hetao Basin base on ensemble learning optimization[J]. Earth Science Frontiers, 2024, 31(3): 371-380.

Figures/Tables 11

References 58

[1]	PODGORSKI J, BERG M. Global threat of arsenic in groundwater[J]. Science, 2020, 368(6493): 845-850. DOI PMID
[2]	WORLD HEALTH ORGANIZATION. Guidelines for Drinking-water Quality[S]. 4th ed. Geneva: World Health Organization, 2011.
[3]	OREMLAND R S, STOLZ J F. The ecology of arsenic[J]. Science, 2003, 300(5621): 939-944. DOI PMID
[4]	JIA Y F, GUO H M, JIANG Y X, et al. Hydrogeochemical zonation and its implication for arsenic mobilization in deep groundwaters near alluvial fans in the Hetao Basin, Inner Mongolia[J]. Journal of Hydrology, 2014, 518(Part C): 410-420.
[5]	SMEDLEY P L, KINNIBURGH D. A review of the source, behaviour and distribution of arsenic in natural waters[J]. Applied Geochemistry, 2002, 17(5): 517-568.
[6]	曹文庚, 董秋瑶, 谭俊, 等. 河套盆地晚更新世以来黄河改道对高砷地下水分布的控制机制[J]. 南水北调与水利科技, 2021, 19(1): 140-150.
[7]	金银龙, 梁超轲, 何公理, 等. 中国地方性砷中毒分布调查(总报告)[J]. 卫生研究, 2003, 32(6): 519-540.
[8]	SMEDLEY P L, ZHANG M, ZHANG G, et al. Mobilisation of arsenic and other trace elements in fluviolacustrine aquifers of the Huhhot Basin, Inner Mongolia[J]. Applied Geochemistry, 2003, 18(9): 1453-1477.
[9]	郭华明, 唐小惠, 杨素珍, 等. 土著微生物作用下含水层沉积物砷的释放与转化[J]. 现代地质, 2009, 23(1): 86-93.
[10]	高存荣, 刘文波, 冯翠娥, 等. 干旱、半干旱地区高砷地下水形成机理研究: 以中国内蒙古河套平原为例[J]. 地学前缘, 2014, 21(4): 13-29.
[11]	CAO W G, GUO H M, ZHANG Y L, et al. Controls of paleochannels on groundwater arsenic distribution in shallow aquifers of alluvial plain in the Hetao Basin, China[J]. The Science of the Total Environment, 2018, 613/614: 958-968.
[12]	GUO H M, LI X M, XIU W, et al. Controls of organic matter bioreactivity on arsenic mobility in shallow aquifers of the Hetao Basin, P. R. China[J]. Journal of Hydrology, 2019, 571: 448-459.
[13]	张庆卜. 国家级地下水位监测数据分析研究: 以民勤盆地为例[D]. 北京: 中国地质大学(北京), 2020.
[14]	CHOWDHURY M, ALOUANI A, HOSSAIN F. Comparison of ordinary Kriging and artificial neural network for spatial mapping of arsenic contamination of groundwater[J]. Stochastic Environmental Research and Risk Assessment, 2010, 24(1): 1-7.
[15]	LIN Y P, CHENG B Y, CHU H J, et al. Assessing how heavy metal pollution and human activity are related by using logistic regression and Kriging methods[J]. Geoderma, 2011, 163(3/4): 275-282.
[16]	AHN J S, CHO Y C. Predicting natural arsenic contamination of bedrock groundwater for a local region in Korea and its application[J]. Environmental Earth Sciences, 2013, 68(7): 2123-2132.
[17]	WINKEL L, BERG M, AMINI M, et al. Predicting groundwater arsenic contamination in Southeast Asia from surface parameters[J]. Nature Geoscience, 2008, 1(8): 536-542.
[18]	TAN Z, YANG Q, ZHENG Y. Machine learning models of groundwater arsenic spatial distribution in Bangladesh: influence of Holocene sediment depositional history[J]. Environmental Science & Technology, 2020, 54(15): 9454-9463.
[19]	TWARAKAVI N K C, MISRA D, BANDOPADHYAY S. Prediction of arsenic in bedrock derived stream sediments at a gold mine site under conditions of sparse data[J]. Natural Resources Research, 2006, 15(1): 15-26.
[20]	CHO K H, STHIANNOPKAO S, PACHEPSKY Y A, et al. Prediction of contamination potential of groundwater arsenic in Cambodia, Laos, and Thailand using artificial neural network[J]. Water Research, 2011, 45(17): 5535-5544. DOI PMID
[21]	LOMBARD M A, BRYAN M S, JONES D K, et al. Machine learning models of arsenic in private wells throughout the conterminous United States as a tool for exposure assessment in human health studies[J]. Environmental Science & Technology, 2021, 55(8): 5012-5023.
[22]	FU Y, CAO W G, PAN D, et al. Changes of groundwater arsenic risk in different seasons in Hetao Basin based on machine learning model[J]. The Science of the Total Environment, 2022, 817: 153058.
[23]	CAO H L, XIE X J, WANG Y X, et al. The interactive natural drivers of global geogenic arsenic contamination of groundwater[J]. Journal of Hydrology, 2021, 597: 126214.
[24]	BUI D T, KHOSRAVI K, TIEFENBACHER J, et al. Improving prediction of water quality indices using novel hybrid machine-learning algorithms[J]. The Science of the Total Environment, 2020, 721: 137612.
[25]	MALLICK J, TALUKDAR S, ALSUBIH M, et al. Integration of statistical models and ensemble machine learning algorithms (MLAs) for developing the novel hybrid groundwater potentiality models: a case study of semi-arid watershed in Saudi Arabia[J]. Geocarto International, 2022, 37(22): 6442-6473.
[26]	CHEN Y, CHEN W, CHANDRA PAL S, et al. Evaluation efficiency of hybrid deep learning algorithms with neural network decision tree and boosting methods for predicting groundwater potential[J]. Geocarto International, 2022, 37(19): 5564-5584.
[27]	WOLPERT D H. Stacked generalization[J]. Neural Networks, 1992, 5(2): 241-259.
[28]	CHATZIMPARMPAS A, MARTINS R M, KUCHER K, et al. StackGenVis: alignment of data, algorithms, and models for stacking ensemble learning using performance metrics[J]. IEEE Transactions on Visualization and Computer Graphics, 2021, 27(2): 1547-1557.
[29]	LEDEZMA A, ALER R, SANCHIS A, et al. GA-stacking: evolutionary stacked generalization[J]. Intelligent Data Analysis, 2010, 14(1): 89-119.
[30]	SUN W, TREVOR B A. stacking ensemble learning framework for annual river ice breakup dates[J]. Journal of Hydrology, 2018, 561: 636-650.
[31]	HU X D, ZHANG H, MEI H B, et al. Landslide susceptibility mapping using the stacking ensemble machine learning method in Lushui, Southwest China[J]. Applied Sciences, 2020, 10(11): 4016.
[32]	TAGHIZADEH R, SCHMIDT K, CHAKAN A A, et al. Improving the spatial prediction of soil organic carbon content in two contrasting climatic regions by stacking machine learning models and rescanning covariate space[J]. Remote Sensing, 2020, 12(7): 1095.
[33]	GU J Y, LIU S G, ZHOU Z Z, et al. A stacking ensemble learning model for monthly rainfall prediction in the Taihu Basin, China[J]. Water, 2022, 14(3): 492.
[34]	GUO H M, ZHANG Y, XING L N, et al. Spatial variation in arsenic and fluoride concentrations of shallow groundwater from the town of Shahai in the Hetao Basin, Inner Mongolia[J]. Applied Geochemistry, 2012, 27(11): 2187-2196.
[35]	高存荣. 河套平原地下水砷污染机理的探讨[J]. 中国地质灾害与防治学报, 1999(2): 25-32.
[36]	RAPHAËL B, VINCENT C, ERIC R, et al. A review and evaluation of the impacts of climate change on geogenic arsenic in groundwater from fractured bedrock aquifers[J]. Water, Air, & Soil Pollution, 2016, 227(9): 296.
[37]	CHARLET L, POLYA D A. Arsenic in shallow, reducing groundwaters in southern Asia: an environmental health disaster[J]. Elements, 2006, 2(2): 91-96.
[38]	PODGORSKI J E, EQANI S, KHANAM T, et al. Extensive arsenic contamination in high-pH unconfined aquifers in the Indus Valley[J]. Science Advances, 2017, 3(8): e1700935.
[39]	NGUYEN P T, HA D H, NGUYEN H D, et al. Improvement of credal decision trees using ensemble frameworks for groundwater potential modeling[J]. Sustainability, 2020, 12(7): 2622.
[40]	GHOBADI A, CHERAGHI M, SOBHANARDAKANI S, et al. Groundwater quality modeling using a novel hybrid data-intelligence model based on gray wolf optimization algorithm and multi-layer perceptron artificial neural network: a case study in Asadabad Plain, Hamedan, Iran[J]. Environmental Science and Pollution Research, 2022, 29(6): 8716-8730.
[41]	OSMAN A I A, AHMED A N, HUANG Y F, et al. Past, present and perspective methodology for groundwater modeling-based machine learning approaches[J]. Archives of Computational Methods in Engineering, 2022, 29(6): 3843-3859.
[42]	ALI E B, ABDESLAM T, YOUSSEF B. Groundwater quality forecasting using machine learning algorithms for irrigation purposes[J]. Agricultural Water Management, 2021, 245: 106625.
[43]	SINGHA S, PASUPULETI S, SINGHA S S, et al. Prediction of groundwater quality using efficient machine learning technique[J]. Chemosphere, 2021, 276: 130265.
[44]	NASIR N, KANSAL A, ALSHALTONE O, et al. Water quality classification using machine learning algorithms[J]. Journal of Water Process Engineering, 2022, 48: 102920.
[45]	MOSAVI A, SAJEDI HOSSEINI F, CHOUBIN B, et al. Ensemble boosting and bagging based machine learning models for groundwater potential prediction[J]. Water Resources Management, 2021, 35(1): 23-37.
[46]	NAGHIBI S A, AHMADI K, DANESHI A. Application of support vector machine, random forest, and genetic algorithm optimized random forest models in groundwater potential mapping[J]. Water Resources Management, 2017, 31(9): 2761-2775.
[47]	HAMMED M M, ALOMAR M K, KHALEEL F, et al. An extra tree regression model for discharge coefficient prediction: novel, practical applications in the hydraulic sector and future research directions[J]. Mathematical Problems in Engineering, 2021, 2021(1): 1-19.
[48]	HUSEN S, KHAMITKAR S, BHALCHANDRA P, et al. Modeling groundwater spring potential of selected geographical area using machine learning algorithms[M]//Applied computer vision and image processing. Singapore: Springer, 2020: 424-432.
[49]	HUANG X, GAO L, CROSBIE R S, et al. Groundwater recharge prediction using linear regression, multi-layer perception network, and deep learning[J]. Water, 2019, 11(9): 1879.
[50]	MAIR A, EL-KADI A I. Logistic regression modeling to assess groundwater vulnerability to contamination in Hawaii, USA[J]. Journal of Contaminant Hydrology, 2013, 153: 1-23. DOI PMID
[51]	WILSON S R, CLOSE M E, ABRAHAM P. Applying linear discriminant analysis to predict groundwater redox conditions conducive to denitrification[J]. Journal of Hydrology, 2018, 556: 611-624.
[52]	MALLIKARJUNA B, SATHISH K, VENKATA KRISHNA P, et al. The effective SVM-based binary prediction of ground water table[J]. Evolutionary Intelligence, 2021, 14(2): 779-787.
[53]	ZHAO J C, JI G X, TIAN Y, et al. Environmental vulnerability assessment for mainland China based on entropy method[J]. Ecological Indicators, 2018, 91: 410-422.
[54]	周志华, 王珏. 机器学习及其应用[M]. 北京: 清华大学出版社, 2007: 63-72.
[55]	GUO H M, TANG X H, YANG S Z, et al. Effect of indigenous bacteria on geochemical behavior of arsenic in aquifer sediments from the Hetao Basin, Inner Mongolia: evidence from sediment incubations[J]. Applied Geochemistry, 2008, 23(12): 3267-3277.
[56]	VAN GEENA A, ZHENG Y, GOODBRED Jr S, et al. Flushing history as a hydrogeological control on the regional distribution of arsenic in shallow groundwater of the Bengal Basin[J]. Environmental Science & Technology, 2008, 42(7): 2283-2288.
[57]	付宇, 曹文庚, 张娟娟. 基于随机森林建模预测河套盆地高砷地下水风险分布[J]. 岩矿测试, 2021, 40(6): 860-870.
[58]	郭华明, 高志鹏, 修伟. 地下水典型氧化还原敏感组分迁移转化的研究热点和趋势[J]. 地学前缘, 2022, 29(3): 64-75. DOI

类别	变量	描述
气候	气温	年均温度(℃)
	降水	年降雨量(mm)
	蒸散(ET)	平均真实蒸散量(mm)
人类活动	排灌渠影响	单位为m
	水位埋深
	水力梯度
沉积环境	Q_3-4地层厚度	全新世(Q₄)和晚更新世(Q₃) 地层的厚度单位为m
	地面标高(DEM)
	黏土层
	黏沙比
水文地质条件	富水性	单位为L/s
土壤理化特征	浅层和深层土壤理化特征	包括土砂分数、土淤泥分数、土黏土分数、土壤有机碳含量、土壤pH值
其他	坡度	单位为(°)
	土地利用	耕地、建筑物、林地、草地、水系
	植被指数(NDVI)	归一化植被指数

类别	变量	描述
气候	气温	年均温度(℃)
	降水	年降雨量(mm)
	蒸散(ET)	平均真实蒸散量(mm)
人类活动	排灌渠影响	单位为m
	水位埋深
	水力梯度
沉积环境	Q_3-4地层厚度	全新世(Q₄)和晚更新世(Q₃) 地层的厚度单位为m
	地面标高(DEM)
	黏土层
	黏沙比
水文地质条件	富水性	单位为L/s
土壤理化特征	浅层和深层土壤理化特征	包括土砂分数、土淤泥分数、土黏土分数、土壤有机碳含量、土壤pH值
其他	坡度	单位为(°)
	土地利用	耕地、建筑物、林地、草地、水系
	植被指数(NDVI)	归一化植被指数

模型性能指标	模型各指标的性能度量
模型性能指标	RF	ET	TreeBag	XGBoost	AdaBoost	GBDT	SVM	LDA	LR	KNN	MLP
AUC	0.770	0.759	0.752	0.694	0.691	0.685	0.720	0.647	0.632	0.603	0.570
Accuracy	0.698	0.680	0.687	0.632	0.632	0.629	0.577	0.604	0.599	0.582	0.549
Precision	0.678	0.657	0.727	0.591	0.670	0.608	0.537	0.604	0.578	0.591	0.597
Recall	0.701	0.703	0.609	0.747	0.622	0.654	0.828	0.630	0.598	0.591	0.421
F1	0.690	0.685	0.663	0.660	0.646	0.656	0.650	0.617	0.588	0.591	0.494
kappa	0.401	0.440	0.375	0.270	0.264	0.289	0.162	0.208	0.197	0.164	0.109

模型性能指标	模型各指标的性能度量
模型性能指标	RF	ET	TreeBag	XGBoost	AdaBoost	GBDT	SVM	LDA	LR	KNN	MLP
AUC	0.770	0.759	0.752	0.694	0.691	0.685	0.720	0.647	0.632	0.603	0.570
Accuracy	0.698	0.680	0.687	0.632	0.632	0.629	0.577	0.604	0.599	0.582	0.549
Precision	0.678	0.657	0.727	0.591	0.670	0.608	0.537	0.604	0.578	0.591	0.597
Recall	0.701	0.703	0.609	0.747	0.622	0.654	0.828	0.630	0.598	0.591	0.421
F1	0.690	0.685	0.663	0.660	0.646	0.656	0.650	0.617	0.588	0.591	0.494
kappa	0.401	0.440	0.375	0.270	0.264	0.289	0.162	0.208	0.197	0.164	0.109

模型	综合评价得分	排名
RF	0.640	1
ET	0.639	2
TreeBag	0.624	3
XGBoost	0.571	4
AdaBoost	0.568	5
GBDT	0.565	6
SVM	0.539	7
LDA	0.528	8
LR	0.510	9
KNN	0.497	10
MLP	0.443	11