-
土壤是陆地生态系统的重要组成部分,可提供维系植物生长的营养物质和微生物生存的家园,是保障人类生存和发展的物质基础[1]. 然而,随着城市化进程的加快,矿产资源开发、金属加工冶炼、化工生产、污水灌溉以及不合理的化肥农药施用等因素导致重金属在农田土壤中不断富集[2],由此造成的粮食减产等问题引起关注. 土壤中重金属主要来源于自然因素和人为因素. 其中,自然因素主要为成土母质,包括土壤自身理化性质的不同所产生的土壤背景值的差异性,同时区域土壤侵蚀程度的不同也会影响成土母质中重金属的释放. 人为因素主要由工业排放、农业活动和交通运输等人类活动等途径产生并在土壤中累积. 重金属的过量累积对土壤孔隙结构、土壤氧化还原环境以及土壤微生物活性产生影响,从而危害整个农田生态安全.
农田土壤与人类生产生活密切相关,一旦受到污染不仅会造成农作物产量和品质的下降,更会对人类健康产生威胁. 中国土壤污染总的超标率为16.1%,其中无机型污染即8种重金属污染为主要类型,占全部超标位点的82.8%[3]. 多项研究表明多种重金属对人体有强烈的致癌致畸毒性:铅中毒会引起人体贫血、便秘、腹痛、呕吐及食欲减退等的临床症状,儿童中毒时会出现智力障碍和行为异常等症状[4]. 长期暴露在镉 (Cd) 污染的环境下会导致肺癌,肾功能发育不良和骨折等人体疾病的高发[5]. 早在20世纪30年代,日本富士县就发生过严重的重金属镉污染稻米引起的数百人骨痛病的土壤污染公害事件[6]. 另外,食用铬 (Cr) 超标的农产品,会引起克汀病、肝癌等人体疾病. 因此,如果农田土壤受到重金属污染,由此产生的食品安全问题将难以忽视,加强农田土壤重金属来源因素的研究是区域重金属风险防控及监测管控的必要条件和重要基础.
目前对于农田土壤重金属污染的研究大多集中在小区域农田,对于大区域污染源问题仍在探索[7]. 近年来,机器学习能够分析土壤特征和重金属浓度大数据集,识别出与重金属浓度密切相关的因素,在处理非线性问题上表现出优势. 目前,已有随机森林、支持向量机、神经网络等多种机器学习方法被广泛应用在土壤重金属预测和监测农田重金属污染方面,随着越来越多的数据积累,这些技术准确度和可信度都将得到提升[8-10]. Catboost模型是一种机器学习算法,能够处理高维度数据并自动选择最重要的特征指标,降低模型的复杂度,在噪声数据或异常值时也能够保持较好的预测能力具有较好的鲁棒性. 相对于其他机器学习算法,Catboost模型的训练速度较快,能够自动处理缺失数据,减少因数据缺失而导致的误差. 因此,在土壤重金属源解析中,Catboost模型是一种具有潜力和优势的机器学习算法.
本文选择中国具有代表性的三大区域,东北三省区域、京津冀和长江经济带作为研究区域. 综述了三大区域的农田重金属研究文献,分析了三大区域重金属的污染特征和潜在生态风险特征,整合与土壤重金属关系密切的辅助因素,通过CatBoost模型分析对农田土壤重金属风险因子进行影响评价,为农田土壤重金属影响因素风险评估提出一种分析方法,以期为我国农田土壤污染防治和环境管理提供一定的评估思路.
基于Catboost算法的中国典型农业区重金属污染特征及影响因素分析
Analysis of heavy metal pollution characteristics and influencing factors in China's typical agricultural areas based on Catboost algorithm
-
摘要: 土壤是保障人类生存和发展的物质基础,农田土壤与人类生产生活密切相关,农田土壤污染问题值得关注. 本文选取东北三省、京津冀和长江经济带作为研究区域,通过文献调研筛选出2000—2020年发布的400篇相关文献,收集到
2052 表层土壤样本,通过统计学、空间分析和机器学习等方法对其污染特征进行分析. 结果表明,与农用地最严格的筛选值标准相比,3个研究区域Cd污染问题较为突出,东北三省、京津冀和长江经济带的超标率分别为37.5%、34.0%和45.8%,长江经济带也存在Cu污染问题,其超标率为30.6%;进一步将采样点位分为矿区周边点位、市郊区点位和其他农田点位,采用潜在生态风险指数法评价表明,矿区周边样点严重和重风险占比(54.8%)显著高于市郊区样点 (37.2%) 和其他农田样点 (36.6%) . 采用Catboost模型进行影响因子识别发现,矿区周边农田主要受选矿、尾矿暴露等造成的Cd和Hg污染影响,而市郊区农田主要受商业、工业和交通运输业等人类活动带来的Hg污染影响,其他农田土壤中主要受农药化肥的施用以及大型器械化带来的Cd污染影响.-
关键词:
- 农田土壤 /
- 重金属 /
- 污染特征 /
- Catboost模型 /
- 影响因素.
Abstract: Soil is the material foundation for human survival and development. Agricultural soil is closely related to human production and life, and the issue of agricultural soil pollution deserves attention. This study focuses on the research areas of the three northeastern provinces, the Beijing-Tianjin-Hebei region, and the Yangtze River Economic Belt. By conducting literature research, 400 relevant articles published between 2000 and 2020 were selected, and 2,052 topsoil samples were collected. Pollution characteristics were analyzed using statistical, spatial analysis, and machine learning methods.The results indicate that, compared to the strictest screening values for agricultural land, Cd pollution is more prominent in the three research areas, with exceedance rates of 37.5%, 34.0%, and 45.8% for the three regions, respectively. The Yangtze River Economic Belt also exhibits Cu pollution with an exceedance rate of 30.6%. Further classifying the sample sites into mining area adjacent sites, suburban area sites, and other farmland sites, an assessment using the potential ecological risk index (PERI) method shows that mining area adjacent sites have significantly higher proportions of severe and high-risk pollution (54.8%) compared to suburban area points (37.2%) and other agricultural field points (36.6%).The Catboost model was employed to identify influencing factors, revealing that agricultural fields adjacent to mining areas are mainly affected by Cd and Hg pollution caused by ore selection and tailings exposure. Suburban area fields are primarily influenced by Hg pollution resulting from human activities such as commerce, industry, and transportation. Other agricultural fields' soil is mainly affected by Cd pollution due to the application of pesticides and fertilizers and the use of large-scale mechanization. -
表 1 潜在生态风险指数划分标准
Table 1. Criteria for classifying potential ecological risk indices
分级 (单)
Classification (Single)标准
Standard分级 (总)
Classification (Total)标准
Standard低 <40 低度 <150 中 40—80 中度 150—300 较重 80—160 重度 300—600 重 160—320 严重 ≥600 严重 ≥320 表 2 研究区域农田土壤重金属含量特征统计
Table 2. Statistics on the heavy metal content of agricultural soils in the study area
研究区域
Study area重金属
Heavy metals浓度范围/(mg·kg−1)
Concentration range平均浓度/(mg·kg−1)
Average concentration变异系数/%
CV筛选值/(mg·kg−1)
Filter value超标率/%
Excess rate东北三省
区域Cr 1.5—155.0 59.9 48.3 150.0 6.3 Ni 1.5—56.0 26.8 36.5 60.0 4.0 Cu 1.5—92.5 26.9 71.6 50.0 11.4 Cd 0.0—4.4 0.5 163.2 0.3 37.5 Pd 1.2—89.5 26.1 65.1 80.0 8.1 Hg 0.0—1.0 0.2 148.21 0.5 9.1 Zn 0.6—367.1 97.3 78.4 200.0 13.3 As 1.7—24.9 10.6 57.4 25.0 14.3 京津冀 Cr 0.0—99.5 56.1 43.6 150.0 2.1 Ni 0.0—52.0 25.8 41.3 60.0 0.0 Cu 0.0—141.1 31.5 77.5 50.0 14.0 Cd 0.03—1.79 0.3 106.2 0.3 34.0 Pd 0.0—85.9 26.9 59.7 80.0 7.7 京津冀 Hg 0.0—1.2 0.2 141.4 0.5 8.6 Zn 0.0—255.9 87.1 54.0 200.0 5.4 As 0.0—30.0 9.4 57.6 25.0 7.3 长江经济带 Cr 0.8—232.5 76.1 54.2 150.0 6.3 Ni 0.9—89.5 36.6 43.0 60.0 13.3 Cu 1.2—214.5 48.2 83.5 50.0 30.6 Cd 0.0—9.1 0.7 148.4 0.3 45.8 Pd 0.9—163.0 39.7 54.7 80.0 9.8 Hg 0.0—4.3 0.3 169.1 0.5 9.5 Zn 11.8—422.6 122.9 64.8 200.0 12.4 As 0.1—21.3 13.4 63.5 25.0 14.9 表 3 研究区域农田土壤重金属背景浓度(mg·kg−1)及超过背景值的比例 (%)
Table 3. Background concentrations of heavy metals in the study area (mg·kg−1) and the proportion exceeding background values (%)
重金属
Heavy metals东北三省
Heilongjiang-Jilin-Liaoning京津冀
Beijing-Tianjin-Hebei长江经济带上游1
Upstream region of the Yangtze River Economic Belt长江经济带下游
Downstream region of the Yangtze River Economic Belt背景值
Background value比例/%
Proportion背景值
Background value比例/%
Proportion背景值
Background value比例/%
Proportion背景值
Background value比例/%
ProportionCr 51.1 0.7 71.2 29.2 79.8 47.1 67.2 56.0 Cd 0.1 90.9 0.1 94.1 0.3 64.7 0.1 90.9 Ni 21.9 83.3 29.9 96.0 36.7 67.9 28.4 67.5 Cu 17.9 71.4 23.9 65.9 35.1 72.4 23.7 83.8 Hg 0.0 95.5 0.0 94.4 0.1 92.1 0.1 71.5 Pb 23.0 59.5 21.9 58.5 14.8 89.7 11.3 97.4 As 7.0 60.7 10.5 23.8 34.4 69.7 27.2 11.6 Zn 66.9 64.5 81.9 43.2 90.6 44.4 74.8 75.5 1. 长江经济带上游包括云南、四川、贵州、重庆, 下游包括湖北、湖南、安徽、江西、江苏、浙江和上海.
1. The upstream region of Yangtze River Economic Belt encompasses Yunnan, Sichuan, Guizhou and Chongqing. Its downsream includes Hubei, Hunan, Anhui, Jiangxi, Jiangsu, Zhejiang and Shanghai.表 4 重金属浓度与农药化肥施用量Person相关系数 (P<0.05)
Table 4. Correlation coefficient between heavy metal concentration and pesticide and fertilizer application Person (P<0.05)
种类
Species重金属
Heavy metalsCr Ni Cu Cd Pb Zn As Hg 化肥施用量 0.05 0.42 0.14 0.25 0.32 0.19 0.70 0.47 农药施用量 0.13 0.31 0.26 0.14 0.42 0.36 0.68 0.16 表 5 三种样点来源单生态风险指数及单生态风险指数/总生态风险指数 (R)
Table 5. Single ecological risk index [PERI(i)] and R[PERI(i)/PERI]
参数
Parameter区域
Area重金属
Heavy metalsCr Ni Cu Cd Pb Hg Zn As PERI(i) 矿区周边 48.47 3.23 13.73 197.97 14.64 81.23 1.58 9.93 市郊区 1.65 2.09 6.40 95.10 7.37 155.36 0.92 8.72 其他农田 1.77 1.92 4.43 155.48 6.82 106.27 1.10 8.91 R/% 矿区周围 14.77 0.95 3.89 50.50 4.10 22.45 0.45 2.89 市郊区 0.63 0.78 2.49 34.54 2.90 54.87 0.36 3.43 其他农田 0.64 0.69 1.57 54.00 2.40 39.13 0.39 1.19 -
[1] 宋伟, 陈百明, 刘琳. 中国耕地土壤重金属污染概况[J]. 水土保持研究, 2013, 20(2): 293-298. SONG W, CHEN B M, LIU L. Soil heavy metal pollution of cultivated land in China[J]. Research of Soil and Water Conservation, 2013, 20(2): 293-298(in Chinese).
[2] 陈文轩, 李茜, 王珍, 等. 中国农田土壤重金属空间分布特征及污染评价[J]. 环境科学, 2020, 41(6): 2822-2833. CHEN W X, LI Q, WANG Z, et al. Spatial distribution characteristics and pollution evaluation of heavy metals in arable land soil of China[J]. Environmental Science, 2020, 41(6): 2822-2833 (in Chinese).
[3] 环境保护部, 国土资源部. 全国土壤污染状况调查公报[EB] 2023-04-07
[4] 唐翔宇, 朱永官. 土壤中重金属对人体生物有效性的体外试验评估[J]. 环境与健康杂志, 2004, 21(3): 183-185. TANG X Y, ZHU Y G. Advances in in vitro tests in evaluating bioavailability of heavy metals in contaminated soil via oral intake[J]. Journal of Environment and Health, 2004, 21(3): 183-185 (in Chinese).
[5] ZUKOWSKA J, BIZIUK M. Methodological evaluation of method for dietary heavy metal intake[J]. Journal of Food Science, 2008, 73(2): 21-29. [6] 贾晓琳. 区域土壤重金属污染的源汇空间分析和时空模拟研究[D]. 杭州: 浙江大学, 2020. JIA X L. Source-sink spatial analysis and spatio-temporal simulation of soil heavy metal pollution in regional scale[D]. Hangzhou: Zhejiang University, 2020 (in Chinese).
[7] 黄颖. 不同尺度农田土壤重金属污染源解析研究[D]. 杭州: 浙江大学, 2018. HUANG Y. The exploring of heavy metal pollution source apportionment in various scale of agricultural soils[D]. Hangzhou: Zhejiang University, 2018 (in Chinese).
[8] 金昭, 吕建树. 基于机器学习模型的区域土壤重金属空间预测精度比较研究[J]. 地理研究, 2022, 41(6): 1731-1747. JIN Z, LV J S. Comparison of the accuracy of spatial prediction for heavy metals in regional soils based on machine learning models[J]. Geographical Research, 2022, 41(6): 1731-1747 (in Chinese).
[9] 黄煜韬, 施维林, 纪娟, 等. 基于BP神经网络对某电镀厂土壤重金属预测及人体健康风险评价[J]. 生态毒理学报, 2022, 17(2): 278-289. HUANG Y T, SHI W L, JI J, et al. Prediction of soil heavy metals based on BP neural network and assessment of human health risk of an electroplating plant[J]. Asian Journal of Ecotoxicology, 2022, 17(2): 278-289 (in Chinese).
[10] HU Y A, CHENG H F. Application of stochastic models in identification and apportionment of heavy metal pollution sources in the surface soils of a large-scale region[J]. Environmental Science of Technology, 2013, 47(8): 3752-3760. doi: 10.1021/es304310k [11] 国家统计局. 2008年统计年鉴[DB] 2023-04-07
[12] 国家统计局. 2021年国家统计年鉴[DB] 2023-04-07
[13] 中华人民共和国2021年国民经济和社会发展统计公报[EB]. 国家统计局信息公开(stats. gov. cn). 2021. Statistical Bulletin of National Economic and Social Development of the People's Republic of China. [EB]. National Bureau of Statistics information disclosure(in Chinese).
[14] 苗丰顺, 李岩, 高岑, 等. 基于CatBoost算法的糖尿病预测方法[J]. 计算机系统应用, 2019, 28(9): 215-218. MIAO F S, LI Y, GAO C, et al. Diabetes prediction method based on CatBoost algorithm[J]. Computer Systems & Applications, 2019, 28(9): 215-218 (in Chinese).
[15] PROKHORENKOVA L, GUSEV G, VOROBEV, A. et al. CatBoost: Unbiased Boosting with Categorical Features [C]. NIPS'18: Proceedings of the 32nd International Conference on Neural Information Processing Systems , 2018 : 6639–6649 . [16] 郑顺安. 我国典型农田土壤中重金属的转化与迁移特征研究[D]. 杭州: 浙江大学, 2010. ZHENG S A. Studies on the transformation and transport of heavy metals in typical Chinese agricultural soils[D]. Hangzhou: Zhejiang University, 2010 (in Chinese).
[17] 陈曼丽. 重金属在表层土壤中迁移特性研究[D]. 郑州: 郑州大学, 2019. CHEN M L. Study on migration characteristics of heavy metals in surface soil[D]. Zhengzhou: Zhengzhou University, 2019 (in Chinese).
[18] 梁成华, 刘学, 杜立宇, 等. 砷在棕壤中的吸附解吸行为及赋存形态研究[J]. 河南农业科学, 2009, 38(4): 64-68. doi: 10.3969/j.issn.1004-3268.2009.04.017 LIANG C H, LIU X, DU L Y, et al. The studies on the adsorptive-desorptive behaviour of arsenic in brown soil and its form distribution[J]. Journal of Henan Agricultural Sciences, 2009, 38(4): 64-68 (in Chinese). doi: 10.3969/j.issn.1004-3268.2009.04.017
[19] 魏复盛, 杨国治, 蒋德珍, 等. 中国土壤元素背景值基本统计量及其特征[J]. 中国环境监测, 1991, 7(1): 1-6. WEI F S, YANG G Z, JIANG D Z, et al. Basic statistics and characteristics of background values of soil elements in China[J]. Environmental Monitoring in China, 1991, 7(1): 1-6 (in Chinese).
[20] 刘小红, 周东美, 郝秀珍, 等. 九华铜矿重金属环境污染状况研究[J]. 土壤, 2007, 39(4): 551-555. LIU X H, ZHOU D M, HAO X Z, et al. Heavy metal pollution of the environment in Jiuhua copper orefield[J]. Soils, 2007, 39(4): 551-555 (in Chinese).
[21] 中华人民共和国生态环境部, 铅锌冶炼行业危险废物环境管理指南[S]. W020211227380770404835. pdf mee. gov. cn Ministry of Ecology and Environment, PRC, Guide for environmental management of hazardous wastes in lead and zinc smelting industry. [S]. W020211227380770404835. pdf (mee. gov. cn) (in Chinese) .
[22] DONG S W, PAN Y C, GUO H, et al. Identifying influencing factors of agricultural soil heavy metals using a geographical detector: A case study in Shunyi district, China[J]. Land, 2021, 10: 1010. doi: 10.3390/land10101010 [23] 刘孝严, 樊亚男, 刘鹏, 等. 基于文献计量分析的长江经济带农田土壤重金属污染特征[J]. 环境科学, 2022, 43(11): 5169-5179. LIU X Y, FAN Y N, LIU P, et al. Characteristics of heavy metal pollution in farmland soil of the Yangtze River economic belt based on bibliometric analysis[J]. Environmental Science, 2022, 43(11): 5169-5179 (in Chinese).
[24] 张丽娜. 山东省基本农田土壤重金属含量分布特征及其环境容量研究[D]. 济南: 山东师范大学, 2010. ZHANG L N. The research of the distribution characteristics of heavy metals and the environmental capacity of fundamental farmland in Shandong Province[D]. Jinan: Shandong Normal University, 2010 (in Chinese).
[25] 周璐. 重金属Pb2+在砂土中的迁移及吸附特性[D]. 吉林: 东北电力大学, 2019. ZHOU L. Migration characteristic of heavy metal Pb2+ in sand[D]. Jilin: Northeast Dianli University, 2019 (in Chinese).
[26] WANG H Y, GAO B, WANG S S, et al. Removal of Pb(II), Cu(II), and Cd(II) from aqueous solutions by biochar derived from KMnO4 treated hickory wood[J]. Bioresource Technology, 2015, 197: 356-362. doi: 10.1016/j.biortech.2015.08.132 [27] 曾兴. 黏土屏障重金属迁移离心模拟相似性及击穿时间评估方法[D]. 杭州: 浙江大学, 2015. ZENG X. Similitude for centrifuge modelling of heavy metal migration in clay barrier and method for evaluating breakthrough time[D]. Hangzhou: Zhejiang University, 2015 (in Chinese).
[28] 刘闯. 重金属Pb2+污染下红黏土变形特性及微观结构变化规律[D]. 桂林: 桂林理工大学, 2022. LIU C. Microstructure change and deformation characteristics of red clay pollution by heavy metal Pb2+[D]. Guilin: Guilin University of Technology, 2022 (in Chinese).
[29] 李懿. 区域土壤重金属污染风险评价、驱动因子与管控策略研究[D]. 杭州: 浙江大学, 2022. LI Y. Risk evaluation, driving factors and management of soil contaminated by heavy metal in regional area[D]. Hangzhou: Zhejiang University, 2022 (in Chinese).
[30] 苏全龙. 城市土壤重金属污染特征及影响因素研究: 以南京市为例[D]. 南京: 南京大学, 2017. SU Q L. Study on characteristics and influencing factors of heavy metal pollution in urban soils-a case study of Nanjing city[D]. Nanjing: Nanjing University, 2017 (in Chinese).
[31] 李婧, 周艳文, 陈森, 等. 我国土壤镉污染现状、危害及其治理方法综述[J]. 安徽农学通报, 2015, 21(24): 104-107. LI J, ZHOU Y W, CHEN S, et al. Actualities, damage and management of soil cadmium pollution in China[J]. Anhui Agricultural Science Bulletin, 2015, 21(24): 104-107 (in Chinese).
[32] 罗耀东. 嘉陵江上游典型县域耕地土壤重金属污染特征分析及风险评价[D]. 兰州: 西北师范大学, 2021. LUO Y D. Characteristic analysis and risk assessment of heavy metal pollution in cultivated soil of a typical county in the upstream of the Jialing River[D]. Lanzhou: Northwest Normal University, 2021 (in Chinese).