基于机器学习的重金属毒性及生态风险预测  被引量:1

Prediction of heavy metal toxicity and ecological risk based on machine learning methods

在线阅读下载全文

作  者:李国锋 于金秋 王宏 池海峰 林姗娜 蔡超[1] LI Guo-feng;YU Jin-qiu;WANG Hong;CHI Hai-feng;LIN Shan-na;CAI Chao(State Key Laboratory of Urban Environment and Health,Institute of Urban Environment,Chinese Academy of Sciences,Xiamen 361021,China;University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区:[1]中国科学院城市环境研究所,城市环境与健康重点实验室,福建厦门361021 [2]中国科学院大学,北京100049

出  处:《中国环境科学》2024年第12期7001-7010,共10页China Environmental Science

基  金:国家重点研发计划项目(2023YFC3709700);福建省中科院STS计划配套项目(2022T3014);福建省科技计划项目(2024N0027)。

摘  要:以土壤典型重金属镉(Cd),铜(Cu),铅(Pb)和锌(Zn)为研究对象,蚯蚓为土壤模式生物,采用文献法搜集已发表论文中重金属对蚯蚓繁殖的半数有效浓度(EC_(50))与所对应的土壤理化性质数据共113组,分析不同数据间关联性,揭示土壤理化因子对重金属生物毒性的影响规律.利用随机森林(RF),梯度提升决策树(GBDT),极限梯度提升(XGBoost),K近临(KNN)和支持向量机(SVR)5种机器学习算法构建机器学习模型,研选最佳模型并开展我国土壤重金属潜在生态风险阈值预测.结果表明,重金属在不同类型土壤中毒性存在显著差异,重金属对蚯蚓的繁殖毒性强弱趋势表现为Cd>Cu>Pb≈Zn.不同土壤理化性质对重金属生物毒性的影响规律不同,其中土壤pH值是影响重金属Pb和Cd的主要因素,对重金属蚯蚓繁殖毒性变化的贡献率分别为57.2%和69.0%;阳离子交换量和有机质含量则分别是重金属Cu和Zn生物毒性的主要影响因子.从模型拟合优度和预测精度对比分析基于土壤理化因子构建的重金属生物毒性机器预测模型的性能,XGBoost模型对Cd,Cu和Zn的生物毒性预测表现较好,而RF模型对Pb的生物毒性预测更准确,训练集和测试集的R^(2)分别达0.939和0.886.利用研选的重金属生物毒性预测模型开展我国34省土壤中重金属生态风险阈值预测,结果发现不同区域土壤潜在生态风险存在明显差异.研究结果可为基于土壤理化性状的重金属生态毒性和潜在生态风险的准确预测与合理评估提供了新的策略.This study focused on the toxicity of typical heavy metals in soil,including cadmium(Cd),copper(Cu),lead(Pb),and zinc(Zn),and summarized their effects on the model organisms,earthworms.A total of 113 datasets encompassing the median effective concentration(EC_(50)) of heavy metals on earthworm reproduction,along with corresponding soil physicochemical properties,were compiled from the published literature.The correlation between various datasets was analyzed to reveal the influence of soil physicochemical factors on the biotoxicity of heavy metals.Five machine learning algorithms,including Random Forest(RF),Gradient Boosting Decision Tree(GBDT),Extreme Gradient Boosting(XGBoost),K-Nearest Neighbor(KNN),and Support Vector Regression(SVR),were employed to develop predictive models for biotoxicity of heavy metals based on soil characteristics,ultimately selecting the best-performing model for predicting potential ecological risk thresholds of heavy metals in Chinese soils.The results indicate significant variation in heavy metal toxicity across different soils,with the toxicity trend for earthworm reproduction ranking as follows:indicate significant variation in heavy metal toxicity across different soil types,with the toxicity ranking for earthworm reproduction as Cd>Cu>Pb≈Zn.The effects of soil physicochemical properties on heavy metal toxicity varies depending on the specific heavy metal.Specifically,soil pH emerged as a key factor influencing the toxicity of Pb and Cd,contributing 57.2% and 69.0% respectively,while cation exchange capacity and organic matter content were found to be the primary influencing factors for the bio-toxicity of Cu and Zn.The performance of the machine prediction models for biological toxicity of heavy metals based on soil physicochemical factors was compared and analyzed in terms of model fit and prediction accuracy.Among the predictive models,the XGBoost model performed well for predicting the bio-toxicity of Cd,Cu,and Zn,while the RF model demonstrated higher accuracy in predicting

关 键 词:机器学习 土壤 重金属 理化性质 生物毒性 生态风险评价 

分 类 号:X53[环境科学与工程—环境工程]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象