基于机器学习的大洋玄武岩构造环境判别研究

徐堃; 关馨儿; 吕豪哲; 赵霄; 热则耶·如孜; 陈艳虹

doi:10.16562/j.cnki.0256-1492.2023041101

基于机器学习的大洋玄武岩构造环境判别研究

1.
中国地质大学（北京）海洋学院，北京 100083

2.
中国地质大学（北京）海洋矿产与极地地质教育部重点实验室，北京 100083

基金项目: 中国地质大学（北京）创新创业训练计划项目（S202211415138）；中央高校基本科研业务费专项资金（2652021007）

详细信息

作者简介: 徐堃（2001—），女，本科，海洋科学专业，E-mail：1011201201@email.cugb.edu.cn

通讯作者: 陈艳虹（1990—），女，博士，主要从事大洋岩石圈的岩石学、地球化学和全球构造研究，E-mail：chenyh@cugb.edu.cn

中图分类号: P736.3

收稿日期: 2023-04-11

修回日期: 2023-06-01

刊出日期: 2024-08-28

Tectonic discrimination of oceanic basalt by machine learning

1.
School of Ocean Sciences, China University of Geosciences, Beijing 100083, China

2.
Key Laboratory of Marine Mineral Resources and Polar Geology, Ministry of Education, China University of Geosciences, Beijing 100083, China

More Information

Corresponding author: CHEN Yanhong, chenyh@cugb.edu.cn

Received Date: 11 April 2023

Revised Date: 01 June 2023

Publish Date: 28 August 2024

摘要

摘要:
玄武岩的地球化学成分与其产出构造环境密切相关，是研究地球深部物质组成与动力学过程的重要岩石。为了判别玄武岩形成的构造环境，前人根据玄武岩的地球化学特征建立了一系列构造判别图，然而这些判别图仅限于二维或三维判别。随着全球玄武岩样品地球化学数据的爆发性增长，这些构造判别图逐渐暴露出其局限性强、准确率较低的缺点。在地学与大数据结合发展的背景下，利用机器学习方法有利于更全面和深入分析数据，建立高准确率和高效率的构造环境判别模型。因此，本文利用GEOROC和PetDB数据库，经过一系列数据下载、处理等步骤，建立了全球现代大洋玄武岩数据集。通过支持向量机（SVM）和随机森林（RF）机器学习算法，训练出高准确率的高维判别模型。本文分析了不同机器学习算法和不同地球化学成分数据集对现代大洋玄武岩构造环境判别的影响，并将各个判别模型应用于蛇绿岩数据当中，探讨机器学习模型在判别古老大洋岩石圈（蛇绿岩）形成构造环境下的应用前景。这项工作为大洋玄武岩形成的构造环境判别提供了更高维度的判别手段，是大数据时代下机器学习如何在地球科学领域应用的一次有益尝试。
- 地球化学 /
- 构造环境 /
- 机器学习 /
- 大洋玄武岩
Abstract:
The geochemical composition of basalt is closely related to the tectonic setting of the formation, thus basalt is an important window for viewing the deep Earth and the composition and geodynamic processes. To discriminate the tectonic setting of basalt formation, although a series of tectonic discrimination diagrams have been established based on the geochemical characteristics of basalt, those discrimination diagrams are limited to two-dimensional or three-dimensional data. With the explosive growth of global geochemical data of basalt, these discrimination diagrams show gradually the shortcomings of being local and inaccurate. Therefore, using machine learning methods is beneficial to analyze data multi-dimensionally and comprehensively, and to establish accurate and efficient discriminant models. A global modern oceanic basalt dataset was established by using GEOROC and PetDB databases through a series of steps from data downloading, training, and analyzing. The dataset was trained by the support vector machine (SVM) and random forest (RF) machine learning algorithms and a high-accuracy and high-dimensional discrimination model was built. In addition, the accuracies of different machine-learning algorithms training were analyzed against different geochemical composition datasets of modern oceanic basalt, and the discrimination models were applied to ophiolitic basalt to explore the application of machine learning models for ancient oceanic basalt. This work provided a higher-dimensional approach to discriminate oceanic basalt, and a successful attempt of using machine learning in earth science in the era of the big data.
- geochemistry /
- tectonic setting /
- machine learning /
- oceanic basalt

HTML全文

图 1 基于机器学习的大洋玄武岩构造环境判别总思路流程图

Figure 1.

下载: 全尺寸图片幻灯片

图 2 下载的全球玄武岩数据分布

Figure 2.

下载: 全尺寸图片幻灯片

图 3 数据处理前后数据统计图

Figure 3.

下载: 全尺寸图片幻灯片

图 4 现代大洋玄武岩数据集IAB、MORB、OIB数据在Ti/1000-V图解和Th-Hf/3-Ta图解上的投影

Figure 4.

下载: 全尺寸图片幻灯片

图 5 本文使用的蛇绿岩数据在Ti-V及Th-Hf/3-Ta判别图上的投影图

Figure 5.

下载: 全尺寸图片幻灯片

表 1 SVM、RF算法下现代大洋玄武岩分类模型准确率

Table 1. Accuracy of modern oceanic basalt classification models using SVM and RF algorithms

	Basic_Data	M&TE	ME	TE	RI	Average
SVM	0.94	0.949	0.979	0.952	0.975	0.959
RF	0.997	0.994	0.914	0.993	0.982	0.976

下载: 导出CSV

表 2 蛇绿岩中玄武岩构造环境预测准确率

Table 2. Prediction accuracy of ophiolite using SVM and RF algorithms

		Basic_Data	M&TE	ME	TE	RI	Average
细分类模型	SVM	0.206	0.235	0.059	0.294	0.206	0.200
细分类模型	RF	0.118	0.265	0.088	0.294	0.265	0.206
大类分类模型	SVM	0.765	0.824	0.765	0.794	0.765	0.783
大类分类模型	RF	0.706	0.794	0.706	0.853	0.765	0.765

下载: 导出CSV

表 3 RF算法下各数据集的分类精确率、召回率与F1分数

Table 3. Classification accuracy, recall, and F1 score of each dataset using RF algorithm

	BABB	IAB	MORB	OPB	OIB	数据集
精确率	1	1	0.99	1	0.99	Basic_Data
召回率	0.99	0.99	1	0.96	0.99
F1分数	0.99	0.99	1	0.98	0.99
精确率	1	0.99	0.99	0.98	0.99	M&TE
召回率	0.98	0.99	1	0.92	0.99
F1分数	0.99	0.99	0.99	0.95	0.99
精确率	0.82	0.84	0.9	0.92	0.91	ME
召回率	0.59	0.78	0.97	0.5	0.91
F1分数	0.69	0.81	0.93	0.65	0.91
精确率	0.99	0.99	0.99	0.98	0.99	TE
召回率	0.97	0.99	1	0.93	0.99
F1分数	0.98	0.99	0.99	0.96	0.99
精确率	0.96	0.97	0.99	0.99	0.98	RI
召回率	0.97	0.96	0.99	0.94	0.97
F1分数	0.97	0.96	0.99	0.96	0.96

下载: 导出CSV

表 4 特征重要性度量

Table 4. The feature importance metric

排名	特征	特征重要性度量	排名	特征	特征重要性度量	排名	特征	特征重要性度量
1	Pb²⁰⁸/Pb²⁰⁴	0.0997	19	Gd	0.0182	37	Ba	0.0013
2	Pb²⁰⁶/Pb²⁰⁴	0.0848	20	Pr	0.0157	38	Ni	0.0012
3	Sr⁸⁷/Sr⁸⁶	0.0845	21	La	0.0143	39	Sm	0.0011
4	Nd¹⁴³/Nd¹⁴⁴	0.0826	22	Nd	0.0124	40	Zr	0.0011
5	Ta	0.0689	23	Nb	0.0112	41	FeO^T	0.0011
6	Li	0.0529	24	Dy	0.0108	42	SiO₂	0.001
7	Ga	0.0492	25	Eu	0.0104	43	P₂O₅	0.0009
8	Cs	0.0458	26	Sc	0.0096	44	Al₂O₃	0.0009
9	Lu	0.0443	27	Ce	0.0072	45	K₂O	0.0009
10	Pb	0.0432	28	Co	0.0063	46	K	0.0008
11	Tm	0.0352	29	Cu	0.0057	47	MgO	0.0008
12	Yb	0.03	30	Sr	0.0048	48	CaO	0.0006
13	U	0.0264	31	Zn	0.0046	49	Rb	0.0006
14	Tb	0.0209	32	Th	0.002	50	Y	0.0006
15	Pb²⁰⁷/Pb²⁰⁴	0.0204	33	Ti	0.0017	51	Mg^#	0.0005
16	Er	0.0193	34	Cr	0.0017	52	Na₂O	0.0005
17	Hf	0.0189	35	TiO₂	0.0015	53	P	0.0004
18	Ho	0.0187	36	V	0.0014	54	MnO	0.0004

下载: 导出CSV

参考文献(31)

[1]	White W M. Probing the Earth’s deep interior through geochemistry[J]. Geochemical Perspectives, 2015, 4(2):95-96.
[2]	Doucet L S, Tetley M G, Li Z X, et al. Geochemical fingerprinting of continental and oceanic basalts: A machine learning approach[J]. Earth-Science Reviews, 2022, 233:104192. doi: 10.1016/j.earscirev.2022.104192
[3]	Pearce J A. Role of the sub-continental lithosphere in magma genesis at active continental margins[M]//Hawkesworth C J, Norry M J. Continental Basalts and Mantle Xenoliths. Nantwich, Cheshire: Shiva Publications, 1983: 230-249.
[4]	Pearce J A. Geochemical fingerprinting of oceanic basalts with applications to ophiolite classification and the search for Archean oceanic crust[J]. Lithos, 2008, 100(1-4):14-48. doi: 10.1016/j.lithos.2007.06.016
[5]	Wood D A. The application of a Th-Hf-Ta diagram to problems of tectonomagmatic classification and to establishing the nature of crustal contamination of basaltic lavas of the British Tertiary Volcanic Province[J]. Earth and Planetary Science Letters, 1980, 50(1):11-30. doi: 10.1016/0012-821X(80)90116-8
[6]	Shervais J W. Ti-V plots and the petrogenesis of modern and ophiolitic lavas[J]. Earth and Planetary Science Letters, 1982, 59(1):101-118. doi: 10.1016/0012-821X(82)90120-0
[7]	Pearce J A. Trace element characteristics of lavas from destructive plate boundaries[M]//Thorpe R S. Orogenic Andesites and Related Rocks. Chichester, England: John Wiley and Sons, 1982: 528-548.
[8]	Rollinson H, Pease V. Using Geochemical Data: To Understand Geological Processes[M]. 2nd ed. Cambridge: Cambridge University Press, 2021: 226-278.
[9]	第鹏飞, 王金荣, 张旗, 等. 玄武岩构造环境判别图评估—全体数据研究的启示[J]. 矿物岩石地球化学通报, 2017, 36(6):891-896,879 DI Pengfei, WANG Jinrong, ZHANG Qi, et al. The evaluation of basalt tectonic discrimination diagrams: Constraints on the research of global basalt data[J]. Bulletin of Mineralogy, Petrology and Geochemistry, 2017, 36(6):891-896,879.]
[10]	Vermeesch P. Tectonic discrimination of basalts with classification trees[J]. Geochimica et Cosmochimica Acta, 2006, 70(7):1839-1848. doi: 10.1016/j.gca.2005.12.016
[11]	周永章, 王俊, 左仁广, 等. 地质领域机器学习、深度学习及实现语言[J]. 岩石学报, 2018, 34(11):3173-3178 ZHOU Yongzhang, WANG Jun, ZUO Renguang, et al. Machine learning, deep learning and Python language in field of geology[J]. Acta Petrologica Sinica, 2018, 34(11):3173-3178.]
[12]	Bergen K J, Johnson P A, De Hoop M V, et al. Machine learning for data-driven discovery in solid Earth geoscience[J]. Science, 2019, 363(6433):eaau0323. doi: 10.1126/science.aau0323
[13]	周志华. 机器学习[M]. 北京: 清华大学出版社, 2016 ZHOU Zhihua. Machine Learning[M]. Beijing: Tsinghua University Press, 2016.]
[14]	刘坤, 刘文波. 机器学习与大陆板内玄武岩构造环境判别[J]. 工程技术与管理, 2017, 1(2):188-191 LIU Kun, LIU Wenbo. Machine learning and identification of the tectonic environment of basalt in the continental plate[J]. Engineering Technology & Management, 2017, 1(2):188-191.]
[15]	焦守涛, 周永章, 张旗, 等. 基于GEOROC数据库的全球辉长岩大数据的大地构造环境智能判别研究[J]. 岩石学报, 2018, 34(11):3189-3194 JIAO Shoutao, ZHOU Yongzhang, ZHANG Qi, et al. Study on intelligent discrimination of tectonic settings based on global gabbro data from GEOROC[J]. Acta Petrologica Sinica, 2018, 34(11):3189-3194.]
[16]	任秋兵, 李明超, 李玉琼, 等. 基于全球橄榄石数据的玄武岩构造环境智能判别方法及其验证[J]. 大地构造与成矿学, 2020, 44(2):212-221 REN Qiubing, LI Mingchao, LI Yuqiong, et al. An intelligent method for geochemical discrimination of tectonic settings of basalt based on olivine composition: GWO-SVM method and its verification[J]. Geotectonica et Metallogenia, 2020, 44(2):212-221.]
[17]	Guo P, Yang T, Xu W L, et al. Machine learning reveals source compositions of intraplate basaltic rocks[J]. Geochemistry, Geophysics, Geosystems, 2021, 22(9):e2021GC009946. doi: 10.1029/2021GC009946
[18]	余星. 海底岩石地球化学研究中的"大数据": PetDB及其应用[J]. 地球科学进展, 2014, 29(2):306-314 YU Xing. The big data tool for seabed petrogeochemistry research-PetDB and its application in geoscience[J]. Advances in Earth Science, 2014, 29(2):306-314.]
[19]	葛粲, 汪方跃, 李永东, 等. 基于GEOROC大数据分析地壳厚度地球化学指标[J]. 岩石学报, 2018, 34(11):3179-3188 GE Can, WANG Fangyue, LI Yongdong, et al. Analysis of geochemical indices of crustal thickness based on GEOROC big data[J]. Acta Petrologica Sinica, 2018, 34(11):3179-3188.]
[20]	张晓琴, 程誉莹. 基于随机森林模型的成分数据缺失值填补法[J]. 应用概率统计, 2017, 33(1):102-110 ZHANG Xiaoqin, CHENG Yuying. Imputation of missing values for compositional data based on random forest[J]. Chinese Journal of Applied Probability and Statistics, 2017, 33(1):102-110.]
[21]	朱紫怡, 周飞, 王瑀, 等. 基于机器学习的锆石成因分类研究[J]. 地学前缘, 2022, 29(5):464-475 ZHU Ziyi, ZHOU Fei, WANG Yu, et al. Machine learning-based approach for zircon classification and genesis determination[J]. Earth Science Frontiers, 2022, 29(5):464-475.]
[22]	Breiman L. Using iterated bagging to debias regressions[J]. Machine Learning, 2001, 45(3):261-277. doi: 10.1023/A:1017934522171
[23]	Cortes C, Vapnik V. Support-vector networks[J]. Machine Learning, 1995, 20(3):273-297.
[24]	Pearce J A. Immobile element fingerprinting of ophiolites[J]. Elements, 2014, 10(2):101-108. doi: 10.2113/gselements.10.2.101
[25]	Dai J G, Wang C S, Stern R J, et al. Forearc magmatic evolution during subduction initiation: Insights from an Early Cretaceous Tibetan ophiolite and comparison with the Izu-Bonin-Mariana forearc[J]. GSA Bulletin, 2021, 133(3-4):753-776. doi: 10.1130/B35644.1
[26]	Clarke D B, Cameron B I, Muecke G K, et al. Early Tertiary basalts from the Labrador Sea floor and Davis Strait region[J]. Canadian Journal of Earth Sciences, 1989, 26(5):956-968. doi: 10.1139/e89-077
[27]	Deng H, Peng S B, Polat A, et al. Neoproterozoic IAT intrusion into Mesoproterozoic MOR Miaowan Ophiolite, Yangtze Craton: evidence for evolving tectonic settings[J]. Precambrian Research, 2017, 289:75-94. doi: 10.1016/j.precamres.2016.12.003
[28]	Güneş A, İlbeyli N, Rasimgil S, et al. Petrological and geochemical characteristics of the diabase and metasomatised dikes from the Tekirova ophiolite (SW Anatolia, Turkey): Tectonomagmatic evolution of the southern Neotethys[J]. Geochemistry, 2021, 81(3):125767. doi: 10.1016/j.chemer.2021.125767
[29]	熊庆. 蛇绿岩记录的大洋地幔内熔体迁移过程[J]. 矿物岩石地球化学通报, 2021, 40(5):999-1011 doi: 10.19658/j.issn.1007-2802.2021.40.043 XIONG Qing. Ophiolitic records of melt migration processes in oceanic mantle[J]. Bulletin of Mineralogy, Petrology and Geochemistry, 2021, 40(5):999-1011.] doi: 10.19658/j.issn.1007-2802.2021.40.043
[30]	卢泓宇, 张敏, 刘奕群, 等. 卷积神经网络特征重要性分析及增强特征选择模型[J]. 软件学报, 2017, 28(11):2879-2890 doi: 10.13328/j.cnki.jos.005349 LU Hongyu, ZHANG Min, LIU Yiqun, et al. Convolution neural network feature importance analysis and feature selection enhanced model[J]. Journal of Software, 2017, 28(11):2879-2890.] doi: 10.13328/j.cnki.jos.005349
[31]	赵庆媛, 叶春茂, 鲁耀兵. 基于随机森林的微动特征重要性评估研究[J]. 现代防御技术, 2022, 50(4):124-131 ZHAO Qingyuan, YE Chunmao, LU Yaobing. A micro-motion feature importance evaluation algorithm based on random forest[J]. Modern Defence Technology, 2022, 50(4):124-131.]