Classification and Human Health Risk Prediction of in-situ Leaching Site of Ionic Rare Earth Ore
-
摘要:
目的 建立离子型稀土矿原地浸矿场地分类及人体健康风险预测模型,为稀土矿浸矿场地生态恢复和污染治理提供参考。 方法 以江西省龙南县为研究区,使用人机交互解译方法获取稀土矿原地浸矿场地空间分布信息;建立浸矿场地分类的模糊层次分析模型;结合分类结果,建立浸矿场地人体健康风险预测的随机森林、提升回归树、C5.0决策树及加权集成模型,运用反距离加权插值法对浸矿场地周边人体健康风险概率进行空间制图。 结果 龙南县稀土矿原地浸矿场地主要集中在县域东北部坡度小于25°的山地丘陵地区;浸矿场地面积是利用模糊层次分析模型开展场地分类的最重要变量;浸矿场地类型在场地集中分布区自东北向西南逐渐由第二类(较低风险)过渡至第四类(高风险);随机森林模型对浸矿场地人体健康风险预测的精度高于提升回归树模型和C5.0决策树模型,模型的决定系数为0.744,使用简单的加权集成方法,可进一步提升预测的精度;浸矿场地人体健康风险在场地集中分布区的中部风险较高、东西两侧风险较低;浸矿场地周边约3500 m范围内人体健康风险较高,3500 m之外风险较低。 结论 模糊层次分析模型是离子型稀土矿原地浸矿场地分类的适用模型;随机森林模型是离子型稀土矿原地浸矿场地人体健康风险预测的最优单一模型。 Abstract:Objective Models for classifying and predicting human health risk of in-situ leaching sites of ionic rare earth ore were established for ecological restoration and pollution control of the leaching sites. Method Taking Longnan County, Jiangxi Province as the study area, spatial distribution of in-situ leaching sites of ionic rare earth ore was obtained by using human-computer interactive interpretation. A fuzzy analytic hierarchy process method was established for classifying the leaching sites. Random forest, lifting regression tree, C5.0 decision tree and weighted integration models were used for predicting human health risk probability of the leaching sites. The inverse distance weighted interpolation method was used to map the spatial distribution of human health risk probability around the leaching sites. Results The leaching sites in Longnan County were mainly located in the mountainous and hilly areas with a slope of less than 25° in the northeast of the county. The area of the leaching sites was the most important variable to classify the leaching sites by using the fuzzy analytic hierarchy process. In the concentrated distribution area of the leaching sites, the risk gradually increased from the second type (with a low risk) to the fourth type (with a high risk) from northeast to southwest of the leaching sites. Random forest model (R2 = 0.744) outperformed the lifting regression tree and the C5.0 decision tree models for predicting the human health risk of the leaching sites. A simple weighted integration method can further improve the prediction of human health risk. The risk was high in the middle and low in the east and west of the concentrated distribution area of the leaching sites. The risk was high within about 3 500 m around the leaching sites and low beyond the distance. Conclusion The fuzzy analytic hierarchy process can be used to reasonably classify the leaching sites of ionic rare earth ore. The random forest model was the optimal model for predicting the human health risk of the leaching sites. -
表 1 离子型稀土矿原地浸矿场地分类指标体系及赋值量化
Table 1. Classification indicator system and indicator quantification of in-situ leaching site of ionic rare earth ore
分类指标类型
Classification indicators type分类指标
Classification indicators赋值标准
Standard of value assignment1 2 3 4 5 场地基本特征 场地面积(hm−2) (0, 10] (10, 20] (20, 30] (30, 40] (40, + ∞] 自然地理要素 坡度(°) (0, 2] (2, 6] (6, 15] (15, 25] (25, + ∞] 植被覆盖度 (0.8, 1] (0.6, 0.8] (0.4, 0.6] (0.2, 0.4] (0, 0.2] 土壤质地类型 黏质土 壤质土 砂质土 受体脆弱性特征 周边主要土地利用类型 未利用地、草地 林地 水体、建设用地 所在镇域范围内平均人口密度(人 km‒2) (0, 0.04] (0.04, 0.08] (0.08, 0.12] (0.12, 0.16] (0.16, + ∞] 表 2 离子型稀土矿原地浸矿场地分类指标体系层次结构及模糊一致性矩阵
Table 2. Hierarchical structure and fuzzy consistency matrix of classification indicators system for in-situ leaching site of ionic rare earth ore
目标层A
Target layer A准则层B
Criteria layer B准则层判断矩阵AB
Criteria layer judgment matrix AB指标层C
Indicator layer C指标层判断矩阵BC
Indicator layer judgment matrix BC浸矿场地类型A 场地基本特征B1 $ \left[\begin{array}{c}A\\ B1\\ B2\\ B3\end{array}\begin{array}{c}B1\\ 1\\ 1/2\\ 1/3\end{array}\begin{array}{c}B2\\ 2\\ 1\\ 1/2\end{array}\begin{array}{c}B3\\ 3\\ 2\\ 1\end{array}\right] $ 场地面积C1 $ \left[\begin{array}{c}B1\\ C1\end{array}\begin{array}{c}C1\\ 1\end{array}\right] $ 自然地理要素B2 坡度C2 $ \left[\begin{array}{c}B2\\ C2\\ C3\\ C4\end{array}\begin{array}{c}C2\\ 1\\ 4\\ 2\end{array}\begin{array}{c}C3\\ 1/4\\ 1\\ 1/2\end{array}\begin{array}{c}C4\\ 1/2\\ 2\\ 1\end{array}\right] $ 植被覆盖度C3 土壤质地类型C4 受体脆弱性特征B3 周边主要土地利用类型C5 $ \left[\begin{array}{c}B3\\ C5\\ C6\end{array}\begin{array}{c}C5\\ 1\\ 3\end{array}\begin{array}{c}C6\\ 1/3\\ 1\end{array}\right] $ 所在镇域平均人口密度C6 表 3 离子型稀土矿原地浸矿场地人体健康风险预测指标集
Table 3. Indicators set for predicting human health risk of in-situ leaching site of ionic rare earth ore
指标类型
Indicators type指标名称
Indicators name描述
Description源 浸矿场地类型 基于模糊层次分析模型确定的浸矿场地类型 浸矿场地面积(ha) 浸矿场地占地面积 所在镇域浸矿场地数量(个) 各场地所在镇域范围内浸矿场地数量 途径 距水体距离(m) 浸矿场地内质心点距离周边水体的最近距离 距道路距离(m) 浸矿场地内质心点距离周边道路的最近距离 地形可达性 浸矿场地内质心点所在部位的地形可达性数值 距县级居民点距离(m) 浸矿场地内质心点距离周边县级居民点的最近距离 距乡镇居民点距离(m) 浸矿场地内质心点距离周边乡镇居民点的最近距离 受体 人口密度(人 km–2) 浸矿场地内质心点所在部位的人口密度数值 表 4 离子型稀土矿原地浸矿场地分类指标综合权重
Table 4. Comprehensive weights of classification indicators of in-situ leaching site of ionic rare earth ore
目标层A
Target layer A准则层B
Criteria layer B指标层C
Indicator layer C综合权重
Comprehensive weights准则层B变量
Criteria layer B variables准则层B权重
Criteria layer B weights指标层C变量
Indicator layer C variables指标层C权重
Indicator layer C weights浸矿场地类型A 场地基本特征B1 0.540 场地面积C1 1 0.540 自然地理要素B2 0.297 坡度C2 0.143 0.042 植被覆盖度C3 0.571 0.170 土壤质地类型C4 0.286 0.085 受体脆弱性特征B3 0.163 周边主要土地利用类型C5 0.250 0.041 所在镇域平均人口密度C6 0.750 0.122 表 5 基于不同机器学习模型的离子型稀土矿原地浸矿场地人体健康风险预测性能
Table 5. Performance of different machine learning models for predicting human health risk of in-situ leaching site of ionic rare earth ore
机器学习模型
Machine
learning model平均误差
Mean error
(ME)均方根误差
Root mean
square error (RMSE)决定系数
Coefficient of
determination (R2)随机森林模型 0.052 0.332 0.744 提升回归树模型 0.087 0.451 0.739 C5.0决策树模型 0.099 0.477 0.699 集成机器学习模型 0.041 0.291 0.784 -
[1] 池汝安, 田 君. 风化壳淋积型稀土矿评述[J]. 中国稀土学报, 2007, 25(6): 641 − 650. doi: 10.3321/j.issn:1000-4343.2007.06.001 [2] 李永绣, 周新木, 刘艳珠, 等. 离子吸附型稀土高效提取和分离技术进展[J]. 中国稀土学报, 2012, 30(3): 257 − 264. [3] 郭钟群, 赵 奎, 金解放, 等. 离子型稀土矿环境风险评估及污染治理研究进展[J]. 稀土, 2019, 40(3): 115 − 126. doi: 10.16533/J.CNKI.15-1099/TF.201903015 [4] 刘 琦, 周 芳, 冯 健, 等. 我国稀土资源现状及选矿技术进展[J]. 矿产保护与利用, 2019, 39(5): 76 − 83. [5] Huang X W, Long Z Q, Li H W, et. al. Development of Rare Earth Hydrometallurgy Technology in China[J]. Journal of Rare Earths, 2005, 23(1): 1 − 4. [6] 吴延熹, 王永志. 离子吸附型稀土矿堆法浸出工艺[P]. 中国, CN89102377.1. 1989-4-16. [7] 尹升华, 吴爱祥. 堆浸矿堆溶液渗流规律初探[J]. 矿业研究与开发, 2006, 26(1): 31 − 33 + 36. doi: 10.3969/j.issn.1005-2763.2006.01.012 [8] 赵中波. 离子型稀土矿原地浸析采矿及其推广应用中值得重视的问题[J]. 南方冶金学院学报, 2000, 21(3): 179 − 183. doi: 10.3969/j.issn.2095-3046.2000.03.006 [9] 丁嘉榆. 离子型稀土矿开发的历史回顾−纪念赣州有色冶金研究所建所60周年[J]. 有色金属科学与工程, 2012, 3(4): 14 − 19. [10] 陈 敏, 张大超, 朱清江, 等. 离子型稀土矿山废弃地生态修复研究进展[J]. 中国稀土学报, 2017, 35(4): 461 − 468. [11] 吕铁硬, 郝多虎, 梅军军. 赣南地区离子型稀土矿山地质环境问题评价方法探讨[J]. 河南科技, 2013, 5(10): 180 − 181. doi: 10.3969/j.issn.1003-5168.2013.10.138 [12] 马 伟, 徐素宁, 王润生, 等. 基于证据权法的赣南稀土矿山地质环境评价[J]. 地球学报, 2015, 36(1): 103 − 110. doi: 10.3975/cagsb.2015.01.12 [13] 姚栋伟, 程 莉, 汪 洋, 等. 基于BP神经网络和GIS的矿山地质环境评价方法[J]. 采矿技术, 2016, 16(3): 56 − 58 + 79. doi: 10.3969/j.issn.1671-2900.2016.03.021 [14] 于 扬, 王登红, 田兆雪, 等. 稀土矿区环境调查SMAIMA方法体系、评价模型及其应用−以赣南离子吸附型稀土矿山为例[J]. 地球学报, 2017, 38(3): 335 − 344. doi: 10.3975/cagsb.2017.03.04 [15] 彭 燕, 何国金, 张兆明, 等. 赣南稀土矿开发区生态环境遥感动态监测与评估[J]. 生态学报, 2016, 36(6): 1676 − 1685. [16] 李 春. 原地浸矿新工艺在离子型稀土矿的推广应用[J]. 有色金属科学与工程, 2011, 2(1): 63 − 67. [17] 罗仙平, 钱有军, 梁长利. 从离子型稀土矿浸取液中提取稀土的技术现状与展望[J]. 有色金属科学与工程, 2012, 3(5): 50 − 53 + 59. doi: 10.13264/j.cnki.ysjskx.2012.05.012 [18] 陈道贵. 基于物元分析和组合权重的原地浸矿技术适用性评价模型[J]. 稀土, 2019, 40(4): 32 − 126. [19] Liu F, Zhang G L, Song X D, et al. High-resolution and three-dimensional mapping of soil texture of China[J]. Geoderma, 2020, 361: 114061. doi: 10.1016/j.geoderma.2019.114061 [20] 包黎莉, 秦承志, 朱阿兴. 地形湿度指数算法误差的定量评价[J]. 地理科学进展, 2011, 30(1): 57 − 64. doi: 10.11820/dlkxjz.2011.01.007 [21] 李苗苗, 吴炳方, 颜长珍, 等. 密云水库上游植被覆盖度的遥感估算[J]. 资源科学, 2004, 26(4): 153 − 159. doi: 10.3321/j.issn:1007-7588.2004.04.022 [22] 连启超. 基于模糊层次分析法的建设工程项目的风险分析及对策研究——以北京M项目为例[D]. 西安: 西安工程大学, 2019. [23] 元云丽. 基于模糊层次分析法(FAHP)的建设工程风险管理研究[D]. 重庆: 重庆大学, 2013. [24] 王 鑫, 于东升, 马利霞, 等. 基于万维网大数据的农药场地土壤污染快速预测方法研究[J]. 土壤学报, 2022, 59(3): 709 − 722. doi: 10.11766/trxb202012300343 [25] Kamisńka J A. A random forest partition model for predicting NO2 concentrations from traffic flow and meteorological conditions[J]. Science of the Total Environment, 2019, 651(1): 475 − 483. [26] Cui P, Dai C Y, Zhang J, et al. Assessing the effects of urban morphology parameters on PM2.5 distribution in Northeast China based on gradient boosted regression trees method[J]. Sustainability, 2022, 14(5): 2618. doi: 10.3390/su14052618 [27] Hou L, Wu X G, Wu Z, et al. Pattern identification and risk prediction of domino effect based on data mining methods for accidents occurred in the tank farm[J]. Reliability Engineering & System Safety, 2020, 193: 106646. [28] Breiman L. Random forest[J]. Machine Learning, 2001, 45(1): 5 − 32. [29] Franklin J. The elements of statistical learning: Data mining, inference and prediction[J]. The Mathematical Intelligencer, 2005, 27: 83 − 85. [30] Bhagavathi S M, Thavasimuthu A, Murugesan A, et al. Weather forecasting and prediction using hybrid C5.0 machine learning algorithm[J]. International Journal of Communication Systems, 2021, 34: 24805. [31] Song X D, Wu H Y, Ju B, et al. Pedoclimatic zone-based three-dimensional soil organic carbon mapping in China[J]. Geoderma, 2020: 114145. [32] 彭雨璇, 谢恩泽, 王美艳, 等. 场地土壤污染信息三维空间插值研究进展及展望[J]. 土壤通报, 2021, 52(5): 1244 − 1250. [33] 李恒凯, 熊云飞, 吴立新. 面向对象的离子吸附型稀土开采高分遥感识别方法[J]. 稀土, 2017, 38(4): 38 − 49. [34] 李恒凯, 瓮旭阳. 离子吸附型稀土开采的高分影像识别尺度选择研究[J]. 中国稀土学报, 2020, 38(2): 240 − 250. [35] Wang X, Yu D S, Ma L X, et al. Using big data searching and machine learning to predict human huealth risk probability from pesticide site soils in China[J]. Journal of Environamtal Management, 2022, 320: 115798. [36] Wang M, Li X, Lei M, et al. Human health risk indentification of petrochemical sites based on extreme gradient boosting[J]. Ecotoxicology and Environmental Safety, 2022, 233: 113332. doi: 10.1016/j.ecoenv.2022.113332 -