en
×

分享给微信好友或者朋友圈

使用微信“扫一扫”功能。
作者简介:

杨永,男,1982年生。正高级工程师,在读博士,从事深海资源和地质大数据应用研究。E-mail:yong0913029@163.com。

通讯作者:

何高文,男,1968年生。正高级工程师,从事深海矿产资源成矿理论研究。E-mail:hegaowen@163.com。

参考文献
Arrhenius G O S, Mero J L, Korkish J. 1964. Origin of oceanic manganese minerals. Science, 144(3615): 170~173.
参考文献
Bau M, Schmidt K, Koschinsky A, Hein J, Kuhn T, Usui A. 2014. Discriminating between different genetic types of marine ferro-manganese crusts and nodules based on rare earth elements and yttrium. Chemical Geology, 381(2014): 1~9.
参考文献
Bergen K J, Johnson P A, Hoop M V, Beroza G C. 2019. Machine learning for data-driven discovery in solid Earth geoscience. Science, 363(1299), http: //dx. doi. org/10. 1126/science. auu0323.
参考文献
Bonatti E, Nayudu Y R. 1965. The origin of oceanic manganese nodules on the ocean floor. American Journal of Science, 263: 17~39.
参考文献
Bonatti E, Kraemer T, Rydell H. 1972. Classification and genesis of submarine iron-manganese deposits. Washington D C National Science Foundation. In: Horn D R, ed. Ferromanganese deposits on the ocean floor. Washington D C: National Science Foundation.
参考文献
Boulton G. 2018. The challegnes of a Big Data Earth. Big Earth Data, https: //doi. org/10. 1080/ 20964471. 2017. 1397411.
参考文献
Boyle P, Frean M. 2005. Dependent Gaussian processes. In: Saul L K, Weiss Y, Bottou L, eds. Advances in Neural Information Processing Systems. Cambridge: MIT Press, 217~224.
参考文献
Burns R G, Burns V M. 1977. Mineralogy of manganese nodules. In: Glasby G P, ed. Marine Manganese Deposits. New York: Elsevier, 185~248.
参考文献
Chen J C, Owen R M. 1989. The hydrothermal component in ferromanganese nodules from the Southeast Pacific Ocean. Geochimica et Cosmocimica Acta, 53(6): 1299~1305.
参考文献
Chen Wei, Zhang Shuqi, Li Renwei, Shahabi H. 2018. Performance evaluation of the GIS-based data mining techniques of best-first decision tree, random forest, and naïve Bayes tree for landslide susceptibility modeling. Science of the Total Environment, 644: 1006~1018.
参考文献
Cracknell M J, Reading A M. 2014. Geological mapping using remote sensing data: A comparison of five machine learning algorithms, their response to variations in the spatial distribution of training data and the use of explicit spatial information. Computers & Geosciences, 63: 22~33.
参考文献
Crerar D A, Barnes H L. 1974. Deposition of deep-sea manganese nodules. Geochimica et Cosmochimica Acta, 38: 279~300.
参考文献
Cronan D S. 1977. Deep-sea nodules-distribution and geochemistry. In: Glabsy G P, ed. Marine Manganese Deposits. Amsterdam: Elsvier.
参考文献
Cronan D S. 1992. Minerals in the EEZ. Chapman & Hall, 209.
参考文献
Cronan D S, Tooms J S. 1969. The geochemistry of manganese nodules and associated pelagic deposits from the Pacific and Indian Oceans. Deep-Sea Research, 16: 335~359.
参考文献
Csato L C. 2002. Gaussian processes-iterative sparse approximations. Doctoral dissertation of Aston University.
参考文献
Dempster A P, Laird N M, Rubin D B. 1977. Maximum likelihood from incomplete data via the EM algorithm. Journal of the Royal Statistical Society, Series B (Methodological), 39(1): 1~38.
参考文献
Deng Xianze, He Gaowen, Xu Yue, Liu Yonggang, Wang Fenlian, Zhang Xiaoyu. 2022. Oxic bottom water dominates polymetallic nodule formation around the Caiwei Guyot, northwestern Pacific Ocean. Ore Geology Reviews, 143(2022), 104776.
参考文献
Di Pengfei, Chen Wanfeng, Zhang Qi, Wang Jinrong, Tang Qingyan, Jiao Shoutao. 2018. Comparison of global N-MORB and E-MORB classification schemes. Acta Petorlogica Sinica, 34(2): 264~275 (in Chinese with English abstract).
参考文献
Dutkiewicz A, Judge A, Müller R D. 2020. Environmental predictors of deep-sea polymetallic nodule occurrence in the global ocean. Geology, 48(3): 293~297.
参考文献
Dymond J, Lyle M, Finney B. 1984. Ferromanganese nodules from MANOP sites H, S, and R-Control of mineraligal and chemical composition by multiple accretionary processes. Geochimica et Cosmochimica Acta, 49: 931~949.
参考文献
Feng Shaorong, Xiao Wenjun. 2008. An improved DBSCAN cluster algorithm. Journal of China University of Mining & Technology, 37(1): 105~111 (in Chinese with English abstract).
参考文献
Fewkes R H. 1973. External and internal features of marine manganese nodules as seen with SEM and their implications in nodules origin. In: Morgenstein M, ed. Papers on the Origin and Distribution of Manganese Nodules in Pacific and Prospects for Exploration. Honolulu: Hawaii Institute of Geophysics, 21~29.
参考文献
Glasby G P. 2006. Manganese: Predominant Role of Odules and Crusts. Berlin: Springer-Verlag.
参考文献
Graham J W, Cooper S C. 1959. Biological origin of manganese-rich deposits on the sea floor. Nature, 183: 1050~1051.
参考文献
Greenslate J. 1974. Microoranisms participate in the construction of manganese nodules. Nature, 249: 181~183.
参考文献
Halbach P, Hebisch U, Scherhag C. 1981. Geochemical variations of ferromanganese nodules and crusts from different provinces of the Pacific-Ocean and their genetic-control. Chemical Geology, 31(1-2): 3~17.
参考文献
Han Jiawei, Pei Jian, Kamber M. 2011. Data Mining: Concepts and Techniques. Armstrong: Elsvier.
参考文献
Hein J R, Schwab W C, Davis A S. 1988. Cobalt- and platinum-rich ferromanganese crusts and associated substrate rocks from the Marshall Islands. Marine Geology, 78(1988): 255~283.
参考文献
Hein J R, Schultz M S, Kan J K. 1990. Insular and submarine ferromanganses minerallization of the Tonga-Lau Region. Marine Mining, 9: 305~354.
参考文献
Hein J R, Koschinsky A. 2013a. Deep-ocean Ferromanganese Crusts and Nodules (Vol. 12). Amsterdam: Elsevier.
参考文献
Hein J R, Mizell K, Koschinsky A, Conrad T A. 2013b. Deep-ocean mineral deposits as a source of critical metals for high- and green-technology applications: Comparison with land-based resources. Ore Geology Reviews, 51: 1~14.
参考文献
Hesse R, Schacht U. 2011. Early Diagenesis of Deep-sea Sediments. Amsterdam: Elsevier.
参考文献
Hey T, Tansley S, Tolle K. 2009. The fourth paradigm: Data-intensive scientific discovery. Redmond, Washington: Microsoft Research.
参考文献
Josso P, Pelleter E, Pourret O, Fouquet Y, Etoubleau J, Cheron S, Bollinger C. 2017. A new discrimination scheme for oceanic ferromanganese deposits using high field strength and rare earth elements. Ore Geology Reviews, 87(2017): 3~15.
参考文献
Kagesten G, Fiorentino D, Baumgartner F, Zillen L. 2019. How do continuous high-resolution models of patchy seabed habitats enhance classification schemes. Geosciences, 9: 237.
参考文献
Kaufman L, Rousseeuw P J. 1987. Clustering by means of Medoids. Statistical Data Analysis Based on the L1-Norm and Related Methods, 405~416.
参考文献
Kempler S, Mathews T. 2017. Earth science data analytics: Definitions, techniques and skills. International Journal of Digital Earth, 16(6): 1~8.
参考文献
Lecours V. 2018. Habitat mapping encyclopedia of ecology. Elsevier, 89: 19~30.
参考文献
Lehnert K, Su Y, Langmuir C H, Sarbas B, Nohl U. 2000. A global geochemical database structure for rocks. Geochemistry, Geophysics, Geosystems, 1(11): 1012.
参考文献
Lifshits M. 2012. Lectures on Gaussian Processes. New York: Springer.
参考文献
Li Guoqing, Liu Ying, Pang Lushen. 2018. Handbook of Big Data Analytics and Mining in Earth Science. Beijing: Posts and Telecom Press (in Chinese with English abstract).
参考文献
Li Zhenggang, Li Huaiming, Hein J R, Dong Yanhui, Wang Mingwei, Ren Xiangwen, Wu Zhaocai, Li Xiaohu, Chu Fengyou. 2021. A possible link between seamount sector collapse and manganese nodule occurrence in the abyssal plains, NW Pacific Ocean. Ore Geology Reviews, 138: 104378.
参考文献
Machida S, Fujinaga K, Ishii T, Nakamura K, Hirano N, Kato Y. 2016. Geology and geochemistry of ferromanganese nodules in the Japanese Exclusive Economic Zone around Minamitorishima Island. Geochemical Journal, 50: 1~17.
参考文献
Mackay D J C. 1998. Introduction to Gaussian processes. In: Bishop C M, ed. Neural Networks and Machine Learning. Verlag: Springer, 89~93.
参考文献
McKelvey V E, Wright N A, Bowen R W. 1983. Analysis of the world distribution of metal-rich subsea manganese nodules. Geological Survey Circular, N886.
参考文献
Mero J L. 1965. The Mineral Resources of the Sea. Amsterdam-London-New York: Elsevier, 178~233.
参考文献
Morgan C L. 2000. Resource Estimates of the Clarion-Clipperton Manganese Nodule Deposits. Boca Raton, Florida: CRC Press.
参考文献
Paluszek M, Thomas S. 2019. MATLAB Machine Learning Recipes: A Problem-Solution Approach. New Jersey, USA.
参考文献
Parsons O E. 2020. A Gaussian mixture model approach to classifying response types. In: Bouguila N, Fan W, eds. Mixture Models and Appkications, Unsupervised. Cambridge: Springer.
参考文献
Piper D Z, Williamson M E. 1977. Composition of Pacific Ocean ferromanganese nodules. Marine Geology, 23: 285~303.
参考文献
Price N B, Calvert S E. 1970. Compositional variation in Pacific Ocean ferromanganese nodules and its relationship to sediment accumulation rates. Marine Geology, 9(1970): 145~171.
参考文献
Rasmussen C E, Williams C K I. 2006. Gaussian Processes for Machine Learning. Massachusetts Institute of Technology. London: MIT Press.
参考文献
Ren Jiangbo, Deng Yinan, Lai Peixin, He Gaowen, Wang Fenlian, Yao Huiqiang, Liu Yonggang. 2021. Geochemical characteristics and genesis of the polymetallic nodules in the Pacific survey area. Earth Science Frontiers, 28(2): 412~425 (in Chinese with English abstract).
参考文献
Ren Jiangbo, He Gaowen, Deng Xiguang, Deng Xianze, Yang Yong, Yao Huiqiang, Yang Shengxiong. 2022. Metallogenesis of Co-rich ferromaganese nodules in the northwestern Pacific: Selective enrichment of metallic elements from seawater. Ore Geology Reviews, 143(2022): 104778.
参考文献
Trugman D T, Shearer P M. 2018. Strong correlation between stress drop and peak ground acceleration for recent M1-4 earthquakes in the San Francisco Bay area. Bulletin of the Seismological Society of America, 108: 929~945.
参考文献
Vermeesch P. 2006. Tectonic discrimination of basalts with classification trees. Geochimica et Cosmochimica Acta, 70: 1839~1848.
参考文献
Wang Chengshan, Hazen R M, Cheng Qiuming, Stephenson M H, Zhou Chenghu, Fox P, Sheng Shuzhong, Oberhansli R, Hou Zengqian, Ma Xiaopan, Feng Zhiqiang, Fan Junxuan, Ma Chao, Hu Xiumian, Luo Bin, Wang Juanle. 2021. The deep-time digital earth program: Data-driven discovery in geosciences. National Science Review, 8: nwab027.
参考文献
Xu Dongyu. 2013. Ocean Mineral Geology. Beijing: Maritime Press (in Chinese with English abstract).
参考文献
Yang Yong, He Gaowen, Ma Jinfeng, Yu Zongze, Yao Huiqiang, Deng Xiguang, Liu Fanglan, Wei Zhenquan. 2020. Acoustic quantitative analysis of ferromanganese nodule and cobalt-rich crust distribution areas using EM122 multibeam backscatter data from deep-sea basin to seamount in Western Pacific Ocean. Deep-Sea Research I, 161: 103281.
参考文献
Yao De, Zhang Lijie, Cui Ruyong. 1996. Mineralogy and geochemistry of ferromanganese crusts from Johnston island EEZ. Marine Geology & Quaternary Geology, 16(1): 33~39 (in Chinese with English abstract).
参考文献
Zhang Jianhui. 2007. K-means Cluster Algorithm Research and Application. Wuhan: Wuhan University of Technology Press (in Chinese with English abstract).
参考文献
Zhang Qi, Sun Weidong, Zhao Yong, Yuan Fanglin, Jiao Shoutao, Chen Wanfeng. 2019. New discrimination diagrams for basalts based on big data research. Big Earth Data, 3(1): 45~55.
参考文献
Zhang Tian, Ramakrishnan R, Livny M. 1996. BIRCH: An efficient data clustering databases method for very large databases. Acm Sigmod Record, 25(2): 103~114.
参考文献
Zhao Pengda. 2019. Characteristics and rational utilization of geological big data. Earth Science Frontiers, 26(4): 1~5 (in Chinese with English abstract).
参考文献
Zhou Yongzhang. 2018. Big Data Mining & Machine Learning in Geoscience. Guangzhou: Sun Yat-Sen University Press (in Chinese with English abstract) .
参考文献
Zhou Yongzhang, Zuo Renguang, Liu Gang, Yuan Feng, Mao Xiancheng, Guo Yanjun, Xiao Fan, Liao Jie, Liu Yanpeng. 2021. The great-leap-forward development of mathematical geoscience during 2010-2019: Big data and artificial intelligence algorithm are changing mathematical geoscience. Bulletin of Mineralogy, Petrology, Geochemistry, 40(3): 556~573 (in Chinese with English abstract).
参考文献
Zuo Renguang, Xiong Yihui, Wang Jian, Carranza E J M. 2019. Deep learning and its application in geochemical mapping. Earth-Science Reviews, 192: 1~14.
参考文献
第鹏飞, 陈万峰, 张旗, 王金荣, 汤庆艳, 焦守涛. 2018. 全球N-MORB和E-MORB分类方案对比. 岩石学报, 34(2): 264~275.
参考文献
冯少荣, 肖文俊. 2008. DBSCAN聚类算法的研究与改进. 中国矿业大学学报, 37(1): 105~111.
参考文献
李国庆, 刘莹, 庞禄申. 2018. 地球科学中的大数据分析与挖掘算法手册. 北京: 人民邮电出版社.
参考文献
任江波, 邓义楠, 赖佩欣, 何高文, 王汾连, 姚会强, 刘永刚. 2021. 太平洋调查区多金属结核的地球化学特征和成因. 地学前缘, 28(2): 412~425.
参考文献
许东禹. 2013. 大洋矿产地质学. 北京: 海洋出版社.
参考文献
姚德, 张丽杰, 崔汝勇. 1996. 约翰斯顿岛附近海域铁锰结核矿物学和地球化学研究. 海洋地质与第四纪地质, 16(1): 33~39.
参考文献
张建辉. 2007. K-means聚类算法研究及应用. 武汉: 武汉理工大学出版社.
参考文献
赵鹏大. 2019. 地质大数据特点及其合理开发利用. 地学前缘, 26(4): 1~5.
参考文献
周永章, 张良均, 张奥多, 王俊. 2018. 地球科学大数据挖掘与机器学习. 广州: 中山大学出版社.
参考文献
周永章, 左仁广, 刘刚, 袁峰, 毛先成, 郭艳军, 肖凡, 廖杰, 刘艳鹏. 2021. 数学地球科学跨越发展的十年: 大数据、人工智能算法正在改变地质学. 矿物岩石地球化学通报, 40(3): 556~573.
目录contents

    摘要

    多金属结核类型成因分类是海底矿产资源关注的重要地质问题,诸多学者一直探索利用多金属结核地球化学特征进行多金属结核成因判别。近年来,随着大数据分析方法的应用,为探索利用机器学习技术进行多金属结核地球化学特征进行成因分类提供了很好的思路和方法。本文基于多年调查研究获取的太平洋多金属结核地球化学数据,利用高斯混合模型聚类分析技术,实现了太平洋深海盆地多金属结核成因分类,并对水成型结核进行了进一步判别分析,共划分出成岩型、混合型、水成Ⅰ型和水成Ⅱ型四类成因多金属结核,为太平洋深海找矿突破和资源评价提供重要依据。同时,不同成因类型结核空间预测结果显示,西北太平洋海域是水成Ⅰ型富钴多金属结核的主要分布区域之一,主要分布在马尔库斯—威克海山群、麦哲伦海山群北部、马绍尔海山群和中太平洋海山群西南部的山间盆地,以及附近的皮嘉费他海盆和中太平洋海盆西北部,是未来西太平洋富钴多金属结核资源找矿突破需要关注的关键海域。

    Abstract

    The genetic classification of ferromanganese nodules is an important geological problem in characterizing seabed mineral resources. The discrimination of nodules by geochemical characteristics has been a challenge since the discovery of nodules in the seabed. In recent years, with the application of Big Earth Data mining and the application of machine learning technologies, creative ideas and methods has resulted in the discrimination of nodules using geochemical characteristics. Based on the geochemical data of ferromanganese nodules obtained in Pacific Ocean over the years, the four types of nodules in deep-sea basin, the diagenetic, mixed, hydrogenetic-Ⅰ and hydrogenetic-Ⅱ types, have been identified successfully by Gaussian mixture model cluster analysis technology. Furthermore, the spatial prediction results of different genetic types of nodules show that the northwest Pacific Ocean is one of the main distribution areas of hydrographic-Ⅰ cobalt-rich nodules. This genetic type is mainly distributed in the intermountain basins of Marcus-Wake seamounts, north of the Magellan seamounts, the Marshall seamounts and the southwest of the Central Pacific seamount group, as well as the nearby Pigafetta basin and the northwest of central Pacific basin. These basins are prospective areas for the future exploration of cobalt-rich ferromanganese nodules in the northwest Pacific Ocean.

  • 深海多金属结核富含Co、Ni、Cu、Mn和REEs等战略性金属矿产,是最具资源潜力和开发利用价值的深海矿产资源,在深海盆地、海山、海底高原、活动和非活动洋中脊、大陆边缘等海底地貌单元均有分布(Cronan,1977),其中水深4000~6500 m的深海盆地是多金属结核的主要赋存区域(Hein et al.,2013a2013b)。太平洋深海盆地是全球多金属结核分布的主要区域(Cronan et al.,1969; Cronan,1977; Hein et al.,2013a2013b),其中东太平洋CC区(即克拉里昂(Clarion)-克里帕顿(Clipperton)断裂带之间的区域)是调查研究程度最高和申请矿区最多的区域,中国在该区域拥有两块多金属结核勘探合同区。近年来,国内外学者在西太平洋海山群山间深海盆地陆续发现高丰度的富钴多金属结核(Machida et al.,2016; Yang Yong et al.,2020; Li Zhenggang et al.,2021; 任江波等,2021; Ren Jiangbo et al.,2022; Deng Xianze et al.,2022)。中国北京先驱高技术公司于2019年在西太平洋成功获得面积约7.4万km2的多金属结核矿区,这是世界上第一个在西太平洋获取的多金属结核勘探合同区,引起了世界各国的关注。

  • 自20世纪60年代以来,世界各国学者对太平洋多金属结核空间分布和地球化学特征进行了诸多研究。例如,Mero(1965)通过对太平洋166个多金属集结核取样站位的主微量元素含量数据进行分析,圈划了太平洋多金属结核高Fe、高Mn、高Ni和Cu、高Co分布区及过渡区域。此后,相关学者利用多金属结核Mn、Fe、Co、Ni、Cu含量和Mn/Fe、Fe/Co和Mn/Ni比值对太平洋多金属结核地球化学特征进行了较为详细的分析(Cronan et al.,1969; Price and Calvert,1970; Cronan,1977; Piper et al.,1977; Mckelvey et al.,1983),在太平洋圈划了富Co、富Ni和Cu、富Mn和富Fe多金属结核分布区。

  • 在对多金属结核地球化学特征系统研究和大量数据统计分析的基础上,Halbach et al.(1981)通过对东北和东南太平洋392个结核样品的地球化学特征分析研究,得出了Mn-Fe-(Co、Ni、Cu)的三角图进行结核的成因分类,得到多金属结核水成成因(Hein et al.,1988)、成岩成因(Halbach et al.,1981)和热液成因(Chen et al.,1989)类型结核。Bau et al.(2014)提出利用Ce、Y和Nd元素含量变化来判别成岩成因、水成成因和热液成因类型结核。Josso et al.( 2017)提出了利用高场强元素(如Zr、Ce、Y)与Mn、Fe、Cu、Ni构三角图进行结核成因分类。许东禹(2013)将多金属结核主要分为水成成因型、氧化成岩型、亚氧化成岩型和热液型四种类型。任江波等(2021)对太平洋186件样品多金属结核地球化学和矿相学特征分析后认为,西太平洋多金属结核具有较高的Co含量和低Mn/Fe比值(平均值为1.1),属于典型的水成成因类型,东太平洋多金属结核具有较高的Cu、Ni含量和高Mn/Fe比值(平均值为2.7),为成岩成因和水成成因两种类型。前述研究虽然利用多金属结核地球化学数据进行了结核成因分类和判别分析,具体分析时仅考虑了两个或三个元素含量或比值,故分类方法不够精确,且成因分类的判别边界受当时获取数据量和人为因素影响较大。

  • 近年来,随着地球科学大数据分析模式变革(Kempler et al.,2017; Boulton,2018; Bergen et al.,2019),“数据密集型”的第四科学研究范式转变(Hey et al.,2009; 周永章等,20182021)和“深时数字地球”国际大科学计划(Wang Chengshan et al.,2021)的实施,越来越多学者将机器学习和深度学习等人工智能技术应用于遥感影像分类、滑坡灾害智能预测、矿产资源评价和海底底质监督分类等地球科学领域(Cracknell et al.,2014; Chen Wei et al.,2018; Trugman et al.,2018; Lecours et al.,2018; Zuo Renguang et al.,2019; Kagesten et al.,2019)。特别是Dutkiewicz et al.(2020)利用海洋地形、物理海洋、海洋化学和海洋生物等多学科大数据,借助机器学习技术实现了全球多金属结核概率分布的预测,这为如何利用深海地球科学大数据解决深海矿产资源问题提供了很好的启示。

  • 聚类分析是指将物理或抽象对象的集合分组为由类似的对象组成的多个类的分析过程,其目的是在相似的基础上收集数据来分类,是应用最为广泛的一种无监督学习方法(周永章等,2018李国庆等,2018),主要分为基于距离、密度、层次、网络和模型的五大类技术(Kaufman et al.,1987; Zhang Tian et al.,1996; 张建辉,2007; 冯少荣等,2008; Han Jiawei et al.,2011; Parsons,2020),其中高斯混合模型(Gaussian mixture model,GMM)聚类是近年来发展起来的基于高斯过程(Rasmussen et al.,2006)模型的聚类分析技术,能够实现更高精度的分类和“软分类”(Parsons,2020)。地学研究中的岩石分类、成因分类、矿物分类、构造期次研究、古气候古环境划分等地质问题均可通过聚类分析来解决(赵鹏大,2019),相关学者基于GEOROC和PetDB地球化学大数据(Lehnert et al.,2000),尝试探索决策树、支持向量机等机器学习技术在大洋玄武岩的成因分类和构造背景判别中的应用(Vermeesch,2006; 第鹏飞等,2018; Zhang Qi et al.,2019),这对利用地球化学大数据智能分析解决地质问题具有很好的借鉴作用。

  • 多金属结核成因分类问题是海底矿产资源研究持续探讨的关键地质问题之一,亦可探索利用聚类分析等机器学习实现多金属结核不同成因类型判别和空间预测。本文基于多年调查研究获取的太平洋多金属结核主量金属元素数据,利用高斯混合模型聚类分析技术实现了太平洋多金属聚类分析和成因类型判别,并分析了不同成因类型多金属结核的成矿作用和空间分布特征,对西北太平洋水成成因的富钴多金属结核进行了空间预测。

  • 1 数据获取与处理

  • 1.1 数据获取

  • 本文收集了2365个取样点的多金属结核Mn、Fe、Co、Ni和Cu五种金属地球化学数据(取样点位置如图1所示),其中2174个取样点数据来源于国际海底管理局数据库中心(Center Data Repository,CDR)(http://www.isa.org.jm/),西北太平洋191个取样点数据为2012~2017年期间中国大洋航次调查实测测试分析数据。

  • 1.2 数据分析与处理

  • 1.2.1 数据统计分析

  • 对2365个取样点Mn、Fe、Co、Ni、Cu等五种金属元素含量及Mn/Fe比值、Fe/Co比值、Ni/Cu比值和Mn/Ni比值进行了统计分析。表1分析结果显示,太平洋多金属结核Mn金属含量在4.60%~36.00%之间,平均值为20.28%,中位数为18.92%,标准偏差为5.97%;Fe金属含量在1.69%~27.90%之间,平均值为12.50%,中位数为12.95%,标准偏差为5.13%;Co金属含量在0.03%~1.10%之间,平均值为0.32%,中位数为0.31%,标准偏差为0.14%;Ni金属含量在0.06%~2.00%之间,平均值为0.75%,中位数为0.54%,标准偏差为0.49%;Cu金属含量在0.04%~1.99%之间,平均值为0.58%,中位数为0.33%,标准偏差为0.52%;Mn/Fe比值在0.32~7.93之间,平均值为2.22,中位数为1.38,标准偏差为1.68;Fe/Co比值在10.72~250.00之间,平均值为42.15,中位数为39.21,标准偏差为17.30;Ni/Cu比值在0.05~10.40之间,平均值为1.60,中位数为1.56,标准偏差为0.58;Mn/Ni比值在6.76~134.01之间,平均值为36.29,中位数为32.10,标准偏差为17.77。直方图统计和概率密度估计结果显示(图2),Mn、Fe、Co、Ni和Cu含量均表现为多峰值高斯分布(即正态分布)特征,其中Mn、Ni和Cu含量直方图和概率密度估计曲线形态基本相似,表现为较强相关性,且与Mn/Fe比值直方图和概率密度估计曲线形态亦较为类似;Fe和Co含量直方图大体都表现为两侧两个峰值的高斯分布特征,显示相关性较大。

  • 图1 太平洋海底地形和结核取样位置图

  • Fig.1 Seafloor terrain and geological station of ferromanganese nodules in the Pacific Ocean

  • 图2 太平洋多金属结核地球化学统计直方图和概率密度估计曲线

  • Fig.2 Histograms and probability density estimation curves of geochemistry of ferromanganese nodules in the Pacific Ocean

  • (a)—锰;(b)—铁;(c)—钴;(d)—镍;(e)—铜;(f)—锰铁比;(g)—铁钴比;(h)—镍铜比;(i)—锰镍比

  • (a) —Mn; (b) —Fe; (c) —Co; (d) —Ni; (e) —Cu; (f) —ratio of Mn/Fe; (g) —ratio of Fe/Co; (h) —ratio of Ni/Cu; (i) —ratio of Mn/Ni

  • 表1 太平洋多金属结核主要元素含量统计信息表(2365个取样点)

  • Table1 Statistics of main metal elements of ferromanganese nodules in the Pacific Ocean (2365 samples)

  • 1.2.2 数据网格化处理

  • 为了便于对太平洋多金属结核Mn、Fe、Co、Ni、Cu主要金属地球化学特征进行区域分析,利用ArcGIS地统计分析(geostatistical analyst)工具中的克里格(Empirical Bayesian Kriging)方法,对获取的2365个取样点地球化学数据进行了克里格网格化处理,并对网格化结果进行了误差分析。这些数据点分布极不均匀,为了反映整个太平洋的多金属结核主要金属地球化学宏观变化特征,首先选取网格间距100 km×100 km,搜索半径3000 km对原始数据点进行第一次网格化,并得到网格化预测标准偏差结果,结果显示Mn、Fe、Co、Ni、Cu、Mn/Fe比值和Fe/Co比值网格化预测标准偏差均值分别为4.6、3.6、0.12、0.26、0.31、1.2和21.8。将第一次网格化得到的数据加入到原始数据点中,选取网格间距50 km×50 km,搜索半径500 km进行第二次网格化,对滤波后第二次网格化结果进行深海盆地范围裁剪,得到最终网格化结果。利用网格化结果数据绘制了太平洋深海盆地多金属结核Mn、Fe、Co、Ni、Cu金属含量图与Mn/Fe和Fe/Co比值图(图3),反映了主要金属元素含量及其比值的空间变化特征。

  • 2 高斯混合模型聚类与太平洋结核成因分类

  • 2.1 高斯混合模型聚类技术

  • 2.1.1 聚类分析技术简述

  • 聚类分析是指将物理或抽象对象的集合分组为由类似的对象组成的多个类的分析过程,其目的是在相似的基础上收集数据来分类,一种典型的无监督学习方法(周永章等,2018李国庆等,2018)。聚类分析技术主要分为五大类:① 如K-means和K-medoids算法的基于距离的聚类分析技术(Kaufman et al.,1987; 张建辉,2007);② 如DBSCAN算法的基于密度的聚类分析(冯少荣等,2008);③ 如BIRCH算法的基于层次的聚类分析技术(Zhang Tian et al.,1996);④ 如STING算法的基于网络的聚类分析技术(Han Jiawei et al.,2011);⑤ 如高斯混合模型基于模型的聚类分析技术(Parsons,2020)。

  • 基于模型的聚类方法主要是基于概率模型和神经网络模型的方法,以基于概率模型方法居多。该方法假设数据集是由一系列的概率分布所决定的,给每一个聚簇假定了一个模型,然后在数据集中寻找能够满足某个模型的簇。这个模型可以是数据点在空间中的密度分布函数,它由一系列概率分布决定,也可以是通过基于标准的统计来自动求出聚类的数目。基于模型的聚类分析结果用概率分布来表示,更具可视化并且可以根据这些概率在某个感兴趣的区域重新拟合预测。

  • 2.1.2 高斯混合模型聚类

  • 高斯过程又称正态随机过程,它是一种普遍存在且重要的随机过程,是一组高斯变量的集合。这个集合里面的任意有限个随机变量都服从联合高斯分布。通俗地讲,在任意时间或空间域观察随机过程,若其随机变量的概率分布都满足高斯分布,这个随机过程就是高斯过程(MacKay,1998; Csato,2002; Boyle et al.,2005; Lifshits,2012)。一个高斯过程完全由它的均值函数和协方差函数决定,只要均值函数和协方差函数确定了,这个高斯过程也就确定了。高斯过程最重要的应用是解决分类和回归问题。高斯过程模型属于无参数模型,相对解决的问题复杂度及与其算法比较减少了算法计算量,高斯模型可以解决高维空间(实际上是无限维)的数学问题,结合贝叶斯概率算法,可以实现通过先验概率,推导未知后验输入变量的后验概率,由果推因的概率(Rasmussen et al.,2006)。

  • 图3 太平洋多金属结核主要金属地球化学图

  • Fig.3 Map of main metal content of nodules in the Pacific Ocean

  • (a)—锰;(b)—铁;(c)—钴;(d)—镍;(e)—铜;(f)—锰铁比;(g)—铁钴比

  • (a) —Mn; (b) —Fe; (c) —Co; (d) —Ni; (e) —Cu; (f) —ratio of Mn/Fe; (g) —ratio of Fe/Co

  • 高斯混合模型是用高斯概率密度函数(正态分布曲线)精确地量化事物,它是一个将事物分解为若干的基于高斯概率密度函数形成的模型。该模型是一个无参数隐形混合模型,一个混合模型通常是两个或两个以上的高斯分布的混合(Parsons,2020)。高斯混合模型如公式(1)所示,式中λk为不同变量的加权值,px|θk)为第k个变量的高斯分布,μk为第k个变量的均值,σk2为第k个变量的方差,N为概率分布。

  • pxθk=λkNxμk,σk2
    (1)
  • 在进行高斯混合模型参数估计时,通常采用最大似然法,最大似然估算时采用的算法是最大期望法(Dempster et al.,1977),用以进行变量的后验概率估计。高斯混合模型最大期望参数估计主要包括参数初始化,期望估计,最大值估计和收敛准则确定等几个步骤和过程。高斯混合模型聚类不仅能够得到无监督分类结果,亦能够同时得到属于某一类的后验概率,实现“软”分类,更符合现实认知。

  • 2.2 太平洋多金属结核聚类与成因类型分析

  • 2.2.1 多金属结核地球化学聚类分析

  • 基于网格化处理后的太平洋多金属结核地球化学数据,借助Matlab机器学习工具箱(Paluszek et al.,2019),利用高斯混合模型聚类技术,进行了多金属结核地球化学聚类分析,聚类选择的变量为多金属结核Mn、Fe、Co、Ni、Cu金属含量和Mn/Fe、Fe/Co金属含量比值等7个变量,聚类数为4个,即Idx_1、Idx_2、Idx_3和Idx_4,四个类型结核地球化学元素含量或比值的均值和标准差如表2和表3所示,各类型元素含量协方差矩阵如表4所示,太平洋四种类型多金属结核空间分布和后验概率,以及不同类型结核各金属元素直方图统计特征。具体结果如图4和图5所示。

  • 2.2.2 太平洋多金属结核成因分类

  • 许东禹(2013)在总结前人关于多金属结核分类(Bonatti et al.,1972; Dymond et al.,1984; Hein et al.,1990)的基础上,提出将多金属结核类型主要分为水成型、氧化成岩型、亚氧化成岩型和热液型,其中热液型通常只限于洋中脊和岛弧等热液区及其附近的特殊构造单元中,且钴、镍和铜等微量元素含量很低(Cronan,1992),经济价值低,此次研究中未考虑此结核类型。

  • 表2 四种结核类型主要元素含量均值信息表

  • Table2 Mean values of main metal content of four-type nodule

  • 表3 四种结核类型主要元素含量标准差信息表

  • Table3 Standard deviation values of main metal content of four-type nodule

  • 图4 太平洋多金属结核成因分类结果

  • Fig.4 Results of genetic classification of nodules in the Pacific Ocean

  • (a)—结核成因分类结果;(b~e)—成岩型、水成Ⅰ型、水成Ⅱ型和混合型结核后验概率

  • (a) —results of genetic classification; (b~e) —posterior probabilities of diagenetic, hydrogenetic-Ⅰ and hydrogenetic-Ⅱ and mixed type, respectively

  • 表4 四种结核类型主要元素相关系数矩阵

  • Table4 Correlation coefficient matrixs of main metal content of four-type nodule

  • 本文对水深4000~6500 m之间的深海盆地的多金属结核进行了地球化学聚类分析,在前人研究的基础上,结合聚类分析结果,将太平洋深海盆地多金属结核分为成岩型(Idx_3)、混合型(即亚氧化成岩型,Idx_1)和水成Ⅰ型(Idx_4)和水成Ⅱ型(Idx_2),四种结核类型聚类分析结果的平均后验概率分别为0.96、0.95、0.93和0.94(图4)。各类型结核地球化学及空间分布特征分述如下:

  • (1)成岩型。成岩型(Idx_3)多金属结核Mn金属含量在19.0%~33.1%之间,平均值为28.4%,标准偏差为3.0%;Fe金属含量在4.4%~11.0%之间,平均值为7.4%,标准偏差为1.4%;Co金属含量在0.0~0.25%之间,平均值为0.13%,标准偏差为0.06%;Ni金属含量在0.77%~1.74%之间,平均值为1.36%,标准偏差为0.21%;Cu金属含量在0.30%~1.57%之间,平均值为1.07%,标准偏差为0.30%;Mn/Fe比值在2.48~7.34之间,平均值为4.26,标准偏差为0.87;Fe/Co比值在22.0~213.2之间,平均值为80.5,标准偏差为53.1。分类结果显示(图4),太平洋成岩型多金属结核主要分布在东太平洋克拉里昂断裂带至马尔萨斯断裂带之间的深海盆地,鲍尔海盆和秘鲁海盆,中太平洋海盆局部区域零星分布。

  • (2)混合型。混合型(Idx_1)多金属结核Mn金属含量在12.3%~33.4%之间,平均值为22.7%,标准偏差为3.8%;Fe金属含量在5.2%~20.7%之间,平均值为12.0%,标准偏差为3.4%;Co金属含量在0.07%~0.59%之间,平均值为0.26%,标准偏差为0.09%;Ni金属含量在0.25%~1.77%之间,平均值为0.98%,标准偏差为0.32%;Cu金属含量在0.19%~1.64%之间,平均值为0.73%,标准偏差为0.30%;Mn/Fe比值在0.76~5.73之间,平均值为2.30,标准偏差为0.91;Fe/Co比值在19.7~132.9之间,平均值为55.2,标准偏差为21.6。分类结果显示(图4),太平洋混合型多金属结核分布广泛,主要分布在东起东太平洋克拉里昂断裂带与克拉伯顿断裂带中部深海盆地以北大部区域,一直延伸至默里断裂带,马绍尔群岛与莱茵群岛之间的中太平洋深海盆地大部海域为混合型多金属结核。

  • (3)水成Ⅰ型。水成Ⅰ型(Idx_4)多金属结核Mn金属含量在10.4%~26.7%之间,平均值为17.4%,标准偏差为2.2%;Fe金属含量在9.3%~25.6%之间,平均值为16.3%,标准偏差为3.2%;Co金属含量在0.10%~1.00%之间,平均值为0.36%,标准偏差为0.11%;Ni金属含量在0.00~0.94%之间,平均值为0.49%,标准偏差为0.13%;Cu金属含量在0.09%~0.61%之间,平均值为0.26%,标准偏差为0.09%;Mn/Fe比值在0.21~1.83之间,平均值为1.11,标准偏差为0.29;Fe/Co比值在22.4~83.6之间,平均值为50.5,标准偏差为13.8。分类结果显示(图4),太平洋水成Ⅰ型多金属结核分布亦较为广泛,主要分布在西南太平洋大部海域深海盆地,马尔库斯—威克海山群、麦哲伦海山群、中太平洋海山群、马绍尔海山群附近的深海盆地中。

  • 图5 太平洋深海盆地多金属结核成因判别图

  • Fig.5 Discriminant diagram of nodules in the Pacific Ocean

  • 图中蓝色为成岩型,绿色为混合型,橙色为水成Ⅰ型,黄色为水成Ⅱ型

  • Blue is diagenetic, green is mixed, orange is hydrogenetic-Ⅰ, and yellow is hydrogenetic-Ⅱ

  • (4)水成Ⅱ型。水成Ⅱ型(Idx_2)多金属结核Mn金属含量在5.7%~20.5%之间,平均值为13.8%,标准偏差为2.4%;Fe金属含量在7.8%~21.7%之间,平均值为13.6%,标准偏差为2.3%;Co金属含量在0.08%~0.46%之间,平均值为0.20%,标准偏差为0.06%;Ni金属含量在0.10%~0.86%之间,平均值为0.31%,标准偏差为0.11%;Cu金属含量在0.00~0.47%之间,平均值为0.17%,标准偏差为0.07%;Mn/Fe比值在0.33~1.76之间,平均值为1.0,标准偏差为0.26;Fe/Co比值在43.6~217.9之间,平均值为90.7,标准偏差为23.4。分类结果显示(图4),太平洋水成Ⅱ型多金属结核主要分布在西北太平洋和南大洋靠近陆地的深海盆地中。

  • 3 分析与讨论

  • 3.1 太平洋结核成因分类

  • 多金属结核广泛分布于海底水-沉积物界面上,其主要生长部分,即成矿壳层主要由锰、铁氧化物氢氧化物构成。多金属结核中的金属元素可分为铁、钴和稀土等铁组元素,锰、镍、铜、钡等锰组元素,硅、铝、钾等造岩元素组和磷和钙等生物成因元素组(姚德等,1996)。其中,钴和稀土等铁组元素的成矿富集机制主要取决于铁的成矿作用机制,而镍和铜等锰组元素的成矿富集则与锰成矿作用机制密切相关。

  • 多金属结核成矿物质具有多源性,如陆源、海底火山和热液作用等,但底层海水和沉积物间隙水则是海底结核成矿元素的直接供给源,多金属结核在海底水-沉积物界面主要发生的成矿作用包括化学沉淀作用(如表面接触氧化沉淀、表面吸附和催化氧化沉淀和胶体凝聚沉淀等)(Fewkes,1973; Crerar et al.,1974; Burns et al.,1977)、生物成因作用(如微生物成矿作用和微体生物成矿作用)(Graham et al.,1959; Greenslate,1974)和火山作用(Arrhenius,1964; Bonatti et al.,1965)等三种主要成矿作用。在上述三种成矿作用下,形成了水成型、氧化成岩型、亚氧化成岩型和热液型多金属结核(许东禹,2013),此次成因分类研究中未考虑热液型结核类型。

  • 根据此次聚类分析结果,太平洋深海盆地多金属结核分为成岩型(Idx_3)、混合型(Idx_1)和水成Ⅰ型(Idx_4)和水成Ⅱ型(Idx_2),将分类结果展布到经典的三角判别图(Bonatti et al.,1972; Halbach et al.,1981;图5)上可以看出,成岩型和水成型(Ⅰ和Ⅱ)分类结果与三角判别图结果高度吻合,成岩型结核Mn/Fe比值在2.7~6.8之间变化,平均值为4.1,大于成岩型和水成型结核Mn/Fe比边界值2.5;水成型结核(Ⅰ和Ⅱ)Mn/Fe比值在0.4~1.8之间变化,平均值为1.1,小于成岩型和水成型结核Mn/Fe比边界值2.5;混合型结核Mn/Fe比值在0.9~5.1之间变化,平均值为2.1,接近成岩型和水成型结核Mn/Fe比边界值2.5。

  • 图6 不同类型结核地球化学直方图统计图

  • Fig.6 Histogram of geochemistry of different genetic nodule

  • (a)—锰;(b)—铁;(c)—钴;(d)—镍;(e)—铜;(f)—锰铁比;(g)—铁钴比

  • (a) —Mn; (b) —Fe; (c) —Co; (d) —Ni; (e) —Cu; (f) —ratio of Mn/Fe; (g) —ratio of Fe/Co

  • 可以看出,传统三角判别图难以对水成型结核进行进一步区分,高斯混合模型聚类结果将水成型结核进一步划分为水成Ⅰ型和水成Ⅱ型,它们的Mn/Fe比均值分别为1.07和1.05,仅利用Mn/Fe比值难以区分,但高斯混合模型聚类直方图统计结果显示(图6),水成Ⅱ型结核较水成Ⅰ型结核Mn、Fe、Co、Ni和Cu含量均略低,但Fe/Co比值较高,且差别较大,其中水成Ⅰ型结核Fe/Co比值在22.4~83.6之间,平均值为50.5,标准偏差为13.8,水成Ⅱ型结核Fe/Co比值在43.6~217.9之间,平均值为90.7,标准偏差为23.4,Fe/Co比均值和标准偏差均明显大于水成Ⅰ型结核。

  • 3.2 太平洋不同结核类型空间分布规律

  • 深海多金属结核富含Mn、Co、Ni、Cu和REEs等战略性金属资源,其成矿过程是目前已知最为缓慢的地质过程之一,平均生长速率在10~20 mm/Ma之间(Hein et al.,2013a),大多数结核的形成主要是沉积物-海水界面水成成矿作用和成岩成矿作用的综合结果(Hein et al.,2013b),受深海沉积速率、表层生物生产力、有机碳含量、底流流速和底栖生物活动等诸多因素的影响(Morgan,2000; Glasby,2006; Hesse et al.,2011; Dutkiewicz et al.,2020)。多金属结核在深海盆地、海山、海底高原、活动洋中脊、非活动洋中脊、大陆边缘和近海地形高地等均有分布(Cronan,1977),其中水深4000~6500 m的深海盆地是多金属结核的主要赋存区域(Hein et al.,2013b)。

  • 利用此次聚类分析结果,对太平洋深海盆地多金属结核空间分布规律进行了初步探讨,结果(图4)显示,太平洋深海盆地存在成岩型、混合型、水成Ⅰ型和水成Ⅱ型四类成因多金属结核,其中东太平洋CC区大部和秘鲁海盆为典型的成岩型结核,具有高Mn/Fe比值,高Ni和Cu,低Co的特征,其中Mn/Fe平均比值为4.26,与前人所述的东太平洋CC区的高锰区域Mn/Fe比值4.1相当(Mero,1965),但区域分布有所差异;水成Ⅰ型多金属结核主要分布在西南太平洋和西北太平洋海山群山间盆地及附近深海盆地中,其Mn/Fe平均比值为1.11,与前人所述的西南太平洋至西北太平洋的高钴区域Mn/Fe比值1.3基本相当(Mero,1965)和西太平洋多金属结核Mn/Fe比值1.05相近(任江波等,2021),空间分布区域基本吻合,且与Price et al.(1970) 所述的西南太平洋和西北太平洋低Mn/Fe比值区基本一致。此外,中太平洋海盆为混合成因型结核,其Mn/Fe平均比值为2.30,介于上述成岩型和水成Ⅰ型Mn/Fe比值之间。

  • 不同成因类型多金属结核地球化学直方图统计结果(图6)显示,成岩型(Idx_3)、混合型(Idx_1)与水成Ⅰ型和Ⅱ型(Idx_4和Idx_1)Mn/Fe比值具有明显不同的高斯分布特征,且与Ni高斯分布特征基本一致。成岩型和混合型Mn/Fe比值和Ni、Cu含量变化较大,水成Ⅰ型和Ⅱ型结核Mn/Fe比值非常接近,且变化不大,但从直方图可以看出,水成Ⅰ型较水成Ⅱ型具有富Co、Ni和Mn的特征,Fe含量基本相当,水成Ⅰ型空间分布前人所述的高Co区域大体相吻合(Mero,1965; Cronan et al.,1969),水成Ⅱ型结核主要分布于靠近大陆的西北太平洋和靠近南太平洋群岛附近的深海盆地,其Mn/Fe比值与水成Ⅰ型接近,Mn、Fe、Co、Ni和Cu含量均略低,但Fe/Co比均值和标准偏差明显高于水成Ⅰ型(图6),初步分析认为水成Ⅱ型结核在形成发育过程中可能受陆源碎屑沉积影响较大,导致其主量元素含量较水成Ⅰ型整体偏低,且Fe/Co比值变化较大。

  • 研究结果显示(图4),西北太平洋水成Ⅰ型多金属结核主要分布在153°E~178°W,7°N~26°N范围内,该区域主要位于马尔库斯—威克海山群、麦哲伦海山群北部、马绍尔海山群和中太平洋海山群西南部的山间深海盆地,以及附近的皮嘉费他海盆和中太平洋海盆西北部,中国西太平洋多金属结核合同区即位于该区域,为典型的水成Ⅰ型富钴多金属结核,该区域的富钴多金属结核资源调查研究值得重点关注。

  • 4 结论

  • 基于太平洋多金属结核地球化学数据,利用高斯混合模型机器学习技术,首次成功实现了太平洋深海盆地多金属结核成因类型聚类分析,得到了成岩型、混合型、水成Ⅰ型和水成Ⅱ型四种成因类型多金属结核及其地球化学空间富集特征。

  • 聚类分类结果显示,西北太平洋海域是水成Ⅰ型富钴多金属结核的主要分布区域之一,主要分布在马尔库斯—威克海山群、麦哲伦海山群北部、马绍尔海山群和中太平洋海山群西南部的山间深海盆地,以及附近的皮嘉费他海盆和中太平洋海盆西北部,是未来富钴型多金属结核资源找矿突破的重点区域。

  • 致谢:感谢审稿人对本文提出的宝贵意见和建议。同时,特别感谢北京先驱高技术公司对本文研究的大力支持。

  • 参考文献

    • Arrhenius G O S, Mero J L, Korkish J. 1964. Origin of oceanic manganese minerals. Science, 144(3615): 170~173.

    • Bau M, Schmidt K, Koschinsky A, Hein J, Kuhn T, Usui A. 2014. Discriminating between different genetic types of marine ferro-manganese crusts and nodules based on rare earth elements and yttrium. Chemical Geology, 381(2014): 1~9.

    • Bergen K J, Johnson P A, Hoop M V, Beroza G C. 2019. Machine learning for data-driven discovery in solid Earth geoscience. Science, 363(1299), http: //dx. doi. org/10. 1126/science. auu0323.

    • Bonatti E, Nayudu Y R. 1965. The origin of oceanic manganese nodules on the ocean floor. American Journal of Science, 263: 17~39.

    • Bonatti E, Kraemer T, Rydell H. 1972. Classification and genesis of submarine iron-manganese deposits. Washington D C National Science Foundation. In: Horn D R, ed. Ferromanganese deposits on the ocean floor. Washington D C: National Science Foundation.

    • Boulton G. 2018. The challegnes of a Big Data Earth. Big Earth Data, https: //doi. org/10. 1080/ 20964471. 2017. 1397411.

    • Boyle P, Frean M. 2005. Dependent Gaussian processes. In: Saul L K, Weiss Y, Bottou L, eds. Advances in Neural Information Processing Systems. Cambridge: MIT Press, 217~224.

    • Burns R G, Burns V M. 1977. Mineralogy of manganese nodules. In: Glasby G P, ed. Marine Manganese Deposits. New York: Elsevier, 185~248.

    • Chen J C, Owen R M. 1989. The hydrothermal component in ferromanganese nodules from the Southeast Pacific Ocean. Geochimica et Cosmocimica Acta, 53(6): 1299~1305.

    • Chen Wei, Zhang Shuqi, Li Renwei, Shahabi H. 2018. Performance evaluation of the GIS-based data mining techniques of best-first decision tree, random forest, and naïve Bayes tree for landslide susceptibility modeling. Science of the Total Environment, 644: 1006~1018.

    • Cracknell M J, Reading A M. 2014. Geological mapping using remote sensing data: A comparison of five machine learning algorithms, their response to variations in the spatial distribution of training data and the use of explicit spatial information. Computers & Geosciences, 63: 22~33.

    • Crerar D A, Barnes H L. 1974. Deposition of deep-sea manganese nodules. Geochimica et Cosmochimica Acta, 38: 279~300.

    • Cronan D S. 1977. Deep-sea nodules-distribution and geochemistry. In: Glabsy G P, ed. Marine Manganese Deposits. Amsterdam: Elsvier.

    • Cronan D S. 1992. Minerals in the EEZ. Chapman & Hall, 209.

    • Cronan D S, Tooms J S. 1969. The geochemistry of manganese nodules and associated pelagic deposits from the Pacific and Indian Oceans. Deep-Sea Research, 16: 335~359.

    • Csato L C. 2002. Gaussian processes-iterative sparse approximations. Doctoral dissertation of Aston University.

    • Dempster A P, Laird N M, Rubin D B. 1977. Maximum likelihood from incomplete data via the EM algorithm. Journal of the Royal Statistical Society, Series B (Methodological), 39(1): 1~38.

    • Deng Xianze, He Gaowen, Xu Yue, Liu Yonggang, Wang Fenlian, Zhang Xiaoyu. 2022. Oxic bottom water dominates polymetallic nodule formation around the Caiwei Guyot, northwestern Pacific Ocean. Ore Geology Reviews, 143(2022), 104776.

    • Di Pengfei, Chen Wanfeng, Zhang Qi, Wang Jinrong, Tang Qingyan, Jiao Shoutao. 2018. Comparison of global N-MORB and E-MORB classification schemes. Acta Petorlogica Sinica, 34(2): 264~275 (in Chinese with English abstract).

    • Dutkiewicz A, Judge A, Müller R D. 2020. Environmental predictors of deep-sea polymetallic nodule occurrence in the global ocean. Geology, 48(3): 293~297.

    • Dymond J, Lyle M, Finney B. 1984. Ferromanganese nodules from MANOP sites H, S, and R-Control of mineraligal and chemical composition by multiple accretionary processes. Geochimica et Cosmochimica Acta, 49: 931~949.

    • Feng Shaorong, Xiao Wenjun. 2008. An improved DBSCAN cluster algorithm. Journal of China University of Mining & Technology, 37(1): 105~111 (in Chinese with English abstract).

    • Fewkes R H. 1973. External and internal features of marine manganese nodules as seen with SEM and their implications in nodules origin. In: Morgenstein M, ed. Papers on the Origin and Distribution of Manganese Nodules in Pacific and Prospects for Exploration. Honolulu: Hawaii Institute of Geophysics, 21~29.

    • Glasby G P. 2006. Manganese: Predominant Role of Odules and Crusts. Berlin: Springer-Verlag.

    • Graham J W, Cooper S C. 1959. Biological origin of manganese-rich deposits on the sea floor. Nature, 183: 1050~1051.

    • Greenslate J. 1974. Microoranisms participate in the construction of manganese nodules. Nature, 249: 181~183.

    • Halbach P, Hebisch U, Scherhag C. 1981. Geochemical variations of ferromanganese nodules and crusts from different provinces of the Pacific-Ocean and their genetic-control. Chemical Geology, 31(1-2): 3~17.

    • Han Jiawei, Pei Jian, Kamber M. 2011. Data Mining: Concepts and Techniques. Armstrong: Elsvier.

    • Hein J R, Schwab W C, Davis A S. 1988. Cobalt- and platinum-rich ferromanganese crusts and associated substrate rocks from the Marshall Islands. Marine Geology, 78(1988): 255~283.

    • Hein J R, Schultz M S, Kan J K. 1990. Insular and submarine ferromanganses minerallization of the Tonga-Lau Region. Marine Mining, 9: 305~354.

    • Hein J R, Koschinsky A. 2013a. Deep-ocean Ferromanganese Crusts and Nodules (Vol. 12). Amsterdam: Elsevier.

    • Hein J R, Mizell K, Koschinsky A, Conrad T A. 2013b. Deep-ocean mineral deposits as a source of critical metals for high- and green-technology applications: Comparison with land-based resources. Ore Geology Reviews, 51: 1~14.

    • Hesse R, Schacht U. 2011. Early Diagenesis of Deep-sea Sediments. Amsterdam: Elsevier.

    • Hey T, Tansley S, Tolle K. 2009. The fourth paradigm: Data-intensive scientific discovery. Redmond, Washington: Microsoft Research.

    • Josso P, Pelleter E, Pourret O, Fouquet Y, Etoubleau J, Cheron S, Bollinger C. 2017. A new discrimination scheme for oceanic ferromanganese deposits using high field strength and rare earth elements. Ore Geology Reviews, 87(2017): 3~15.

    • Kagesten G, Fiorentino D, Baumgartner F, Zillen L. 2019. How do continuous high-resolution models of patchy seabed habitats enhance classification schemes. Geosciences, 9: 237.

    • Kaufman L, Rousseeuw P J. 1987. Clustering by means of Medoids. Statistical Data Analysis Based on the L1-Norm and Related Methods, 405~416.

    • Kempler S, Mathews T. 2017. Earth science data analytics: Definitions, techniques and skills. International Journal of Digital Earth, 16(6): 1~8.

    • Lecours V. 2018. Habitat mapping encyclopedia of ecology. Elsevier, 89: 19~30.

    • Lehnert K, Su Y, Langmuir C H, Sarbas B, Nohl U. 2000. A global geochemical database structure for rocks. Geochemistry, Geophysics, Geosystems, 1(11): 1012.

    • Lifshits M. 2012. Lectures on Gaussian Processes. New York: Springer.

    • Li Guoqing, Liu Ying, Pang Lushen. 2018. Handbook of Big Data Analytics and Mining in Earth Science. Beijing: Posts and Telecom Press (in Chinese with English abstract).

    • Li Zhenggang, Li Huaiming, Hein J R, Dong Yanhui, Wang Mingwei, Ren Xiangwen, Wu Zhaocai, Li Xiaohu, Chu Fengyou. 2021. A possible link between seamount sector collapse and manganese nodule occurrence in the abyssal plains, NW Pacific Ocean. Ore Geology Reviews, 138: 104378.

    • Machida S, Fujinaga K, Ishii T, Nakamura K, Hirano N, Kato Y. 2016. Geology and geochemistry of ferromanganese nodules in the Japanese Exclusive Economic Zone around Minamitorishima Island. Geochemical Journal, 50: 1~17.

    • Mackay D J C. 1998. Introduction to Gaussian processes. In: Bishop C M, ed. Neural Networks and Machine Learning. Verlag: Springer, 89~93.

    • McKelvey V E, Wright N A, Bowen R W. 1983. Analysis of the world distribution of metal-rich subsea manganese nodules. Geological Survey Circular, N886.

    • Mero J L. 1965. The Mineral Resources of the Sea. Amsterdam-London-New York: Elsevier, 178~233.

    • Morgan C L. 2000. Resource Estimates of the Clarion-Clipperton Manganese Nodule Deposits. Boca Raton, Florida: CRC Press.

    • Paluszek M, Thomas S. 2019. MATLAB Machine Learning Recipes: A Problem-Solution Approach. New Jersey, USA.

    • Parsons O E. 2020. A Gaussian mixture model approach to classifying response types. In: Bouguila N, Fan W, eds. Mixture Models and Appkications, Unsupervised. Cambridge: Springer.

    • Piper D Z, Williamson M E. 1977. Composition of Pacific Ocean ferromanganese nodules. Marine Geology, 23: 285~303.

    • Price N B, Calvert S E. 1970. Compositional variation in Pacific Ocean ferromanganese nodules and its relationship to sediment accumulation rates. Marine Geology, 9(1970): 145~171.

    • Rasmussen C E, Williams C K I. 2006. Gaussian Processes for Machine Learning. Massachusetts Institute of Technology. London: MIT Press.

    • Ren Jiangbo, Deng Yinan, Lai Peixin, He Gaowen, Wang Fenlian, Yao Huiqiang, Liu Yonggang. 2021. Geochemical characteristics and genesis of the polymetallic nodules in the Pacific survey area. Earth Science Frontiers, 28(2): 412~425 (in Chinese with English abstract).

    • Ren Jiangbo, He Gaowen, Deng Xiguang, Deng Xianze, Yang Yong, Yao Huiqiang, Yang Shengxiong. 2022. Metallogenesis of Co-rich ferromaganese nodules in the northwestern Pacific: Selective enrichment of metallic elements from seawater. Ore Geology Reviews, 143(2022): 104778.

    • Trugman D T, Shearer P M. 2018. Strong correlation between stress drop and peak ground acceleration for recent M1-4 earthquakes in the San Francisco Bay area. Bulletin of the Seismological Society of America, 108: 929~945.

    • Vermeesch P. 2006. Tectonic discrimination of basalts with classification trees. Geochimica et Cosmochimica Acta, 70: 1839~1848.

    • Wang Chengshan, Hazen R M, Cheng Qiuming, Stephenson M H, Zhou Chenghu, Fox P, Sheng Shuzhong, Oberhansli R, Hou Zengqian, Ma Xiaopan, Feng Zhiqiang, Fan Junxuan, Ma Chao, Hu Xiumian, Luo Bin, Wang Juanle. 2021. The deep-time digital earth program: Data-driven discovery in geosciences. National Science Review, 8: nwab027.

    • Xu Dongyu. 2013. Ocean Mineral Geology. Beijing: Maritime Press (in Chinese with English abstract).

    • Yang Yong, He Gaowen, Ma Jinfeng, Yu Zongze, Yao Huiqiang, Deng Xiguang, Liu Fanglan, Wei Zhenquan. 2020. Acoustic quantitative analysis of ferromanganese nodule and cobalt-rich crust distribution areas using EM122 multibeam backscatter data from deep-sea basin to seamount in Western Pacific Ocean. Deep-Sea Research I, 161: 103281.

    • Yao De, Zhang Lijie, Cui Ruyong. 1996. Mineralogy and geochemistry of ferromanganese crusts from Johnston island EEZ. Marine Geology & Quaternary Geology, 16(1): 33~39 (in Chinese with English abstract).

    • Zhang Jianhui. 2007. K-means Cluster Algorithm Research and Application. Wuhan: Wuhan University of Technology Press (in Chinese with English abstract).

    • Zhang Qi, Sun Weidong, Zhao Yong, Yuan Fanglin, Jiao Shoutao, Chen Wanfeng. 2019. New discrimination diagrams for basalts based on big data research. Big Earth Data, 3(1): 45~55.

    • Zhang Tian, Ramakrishnan R, Livny M. 1996. BIRCH: An efficient data clustering databases method for very large databases. Acm Sigmod Record, 25(2): 103~114.

    • Zhao Pengda. 2019. Characteristics and rational utilization of geological big data. Earth Science Frontiers, 26(4): 1~5 (in Chinese with English abstract).

    • Zhou Yongzhang. 2018. Big Data Mining & Machine Learning in Geoscience. Guangzhou: Sun Yat-Sen University Press (in Chinese with English abstract) .

    • Zhou Yongzhang, Zuo Renguang, Liu Gang, Yuan Feng, Mao Xiancheng, Guo Yanjun, Xiao Fan, Liao Jie, Liu Yanpeng. 2021. The great-leap-forward development of mathematical geoscience during 2010-2019: Big data and artificial intelligence algorithm are changing mathematical geoscience. Bulletin of Mineralogy, Petrology, Geochemistry, 40(3): 556~573 (in Chinese with English abstract).

    • Zuo Renguang, Xiong Yihui, Wang Jian, Carranza E J M. 2019. Deep learning and its application in geochemical mapping. Earth-Science Reviews, 192: 1~14.

    • 第鹏飞, 陈万峰, 张旗, 王金荣, 汤庆艳, 焦守涛. 2018. 全球N-MORB和E-MORB分类方案对比. 岩石学报, 34(2): 264~275.

    • 冯少荣, 肖文俊. 2008. DBSCAN聚类算法的研究与改进. 中国矿业大学学报, 37(1): 105~111.

    • 李国庆, 刘莹, 庞禄申. 2018. 地球科学中的大数据分析与挖掘算法手册. 北京: 人民邮电出版社.

    • 任江波, 邓义楠, 赖佩欣, 何高文, 王汾连, 姚会强, 刘永刚. 2021. 太平洋调查区多金属结核的地球化学特征和成因. 地学前缘, 28(2): 412~425.

    • 许东禹. 2013. 大洋矿产地质学. 北京: 海洋出版社.

    • 姚德, 张丽杰, 崔汝勇. 1996. 约翰斯顿岛附近海域铁锰结核矿物学和地球化学研究. 海洋地质与第四纪地质, 16(1): 33~39.

    • 张建辉. 2007. K-means聚类算法研究及应用. 武汉: 武汉理工大学出版社.

    • 赵鹏大. 2019. 地质大数据特点及其合理开发利用. 地学前缘, 26(4): 1~5.

    • 周永章, 张良均, 张奥多, 王俊. 2018. 地球科学大数据挖掘与机器学习. 广州: 中山大学出版社.

    • 周永章, 左仁广, 刘刚, 袁峰, 毛先成, 郭艳军, 肖凡, 廖杰, 刘艳鹏. 2021. 数学地球科学跨越发展的十年: 大数据、人工智能算法正在改变地质学. 矿物岩石地球化学通报, 40(3): 556~573.

  • 参考文献

    • Arrhenius G O S, Mero J L, Korkish J. 1964. Origin of oceanic manganese minerals. Science, 144(3615): 170~173.

    • Bau M, Schmidt K, Koschinsky A, Hein J, Kuhn T, Usui A. 2014. Discriminating between different genetic types of marine ferro-manganese crusts and nodules based on rare earth elements and yttrium. Chemical Geology, 381(2014): 1~9.

    • Bergen K J, Johnson P A, Hoop M V, Beroza G C. 2019. Machine learning for data-driven discovery in solid Earth geoscience. Science, 363(1299), http: //dx. doi. org/10. 1126/science. auu0323.

    • Bonatti E, Nayudu Y R. 1965. The origin of oceanic manganese nodules on the ocean floor. American Journal of Science, 263: 17~39.

    • Bonatti E, Kraemer T, Rydell H. 1972. Classification and genesis of submarine iron-manganese deposits. Washington D C National Science Foundation. In: Horn D R, ed. Ferromanganese deposits on the ocean floor. Washington D C: National Science Foundation.

    • Boulton G. 2018. The challegnes of a Big Data Earth. Big Earth Data, https: //doi. org/10. 1080/ 20964471. 2017. 1397411.

    • Boyle P, Frean M. 2005. Dependent Gaussian processes. In: Saul L K, Weiss Y, Bottou L, eds. Advances in Neural Information Processing Systems. Cambridge: MIT Press, 217~224.

    • Burns R G, Burns V M. 1977. Mineralogy of manganese nodules. In: Glasby G P, ed. Marine Manganese Deposits. New York: Elsevier, 185~248.

    • Chen J C, Owen R M. 1989. The hydrothermal component in ferromanganese nodules from the Southeast Pacific Ocean. Geochimica et Cosmocimica Acta, 53(6): 1299~1305.

    • Chen Wei, Zhang Shuqi, Li Renwei, Shahabi H. 2018. Performance evaluation of the GIS-based data mining techniques of best-first decision tree, random forest, and naïve Bayes tree for landslide susceptibility modeling. Science of the Total Environment, 644: 1006~1018.

    • Cracknell M J, Reading A M. 2014. Geological mapping using remote sensing data: A comparison of five machine learning algorithms, their response to variations in the spatial distribution of training data and the use of explicit spatial information. Computers & Geosciences, 63: 22~33.

    • Crerar D A, Barnes H L. 1974. Deposition of deep-sea manganese nodules. Geochimica et Cosmochimica Acta, 38: 279~300.

    • Cronan D S. 1977. Deep-sea nodules-distribution and geochemistry. In: Glabsy G P, ed. Marine Manganese Deposits. Amsterdam: Elsvier.

    • Cronan D S. 1992. Minerals in the EEZ. Chapman & Hall, 209.

    • Cronan D S, Tooms J S. 1969. The geochemistry of manganese nodules and associated pelagic deposits from the Pacific and Indian Oceans. Deep-Sea Research, 16: 335~359.

    • Csato L C. 2002. Gaussian processes-iterative sparse approximations. Doctoral dissertation of Aston University.

    • Dempster A P, Laird N M, Rubin D B. 1977. Maximum likelihood from incomplete data via the EM algorithm. Journal of the Royal Statistical Society, Series B (Methodological), 39(1): 1~38.

    • Deng Xianze, He Gaowen, Xu Yue, Liu Yonggang, Wang Fenlian, Zhang Xiaoyu. 2022. Oxic bottom water dominates polymetallic nodule formation around the Caiwei Guyot, northwestern Pacific Ocean. Ore Geology Reviews, 143(2022), 104776.

    • Di Pengfei, Chen Wanfeng, Zhang Qi, Wang Jinrong, Tang Qingyan, Jiao Shoutao. 2018. Comparison of global N-MORB and E-MORB classification schemes. Acta Petorlogica Sinica, 34(2): 264~275 (in Chinese with English abstract).

    • Dutkiewicz A, Judge A, Müller R D. 2020. Environmental predictors of deep-sea polymetallic nodule occurrence in the global ocean. Geology, 48(3): 293~297.

    • Dymond J, Lyle M, Finney B. 1984. Ferromanganese nodules from MANOP sites H, S, and R-Control of mineraligal and chemical composition by multiple accretionary processes. Geochimica et Cosmochimica Acta, 49: 931~949.

    • Feng Shaorong, Xiao Wenjun. 2008. An improved DBSCAN cluster algorithm. Journal of China University of Mining & Technology, 37(1): 105~111 (in Chinese with English abstract).

    • Fewkes R H. 1973. External and internal features of marine manganese nodules as seen with SEM and their implications in nodules origin. In: Morgenstein M, ed. Papers on the Origin and Distribution of Manganese Nodules in Pacific and Prospects for Exploration. Honolulu: Hawaii Institute of Geophysics, 21~29.

    • Glasby G P. 2006. Manganese: Predominant Role of Odules and Crusts. Berlin: Springer-Verlag.

    • Graham J W, Cooper S C. 1959. Biological origin of manganese-rich deposits on the sea floor. Nature, 183: 1050~1051.

    • Greenslate J. 1974. Microoranisms participate in the construction of manganese nodules. Nature, 249: 181~183.

    • Halbach P, Hebisch U, Scherhag C. 1981. Geochemical variations of ferromanganese nodules and crusts from different provinces of the Pacific-Ocean and their genetic-control. Chemical Geology, 31(1-2): 3~17.

    • Han Jiawei, Pei Jian, Kamber M. 2011. Data Mining: Concepts and Techniques. Armstrong: Elsvier.

    • Hein J R, Schwab W C, Davis A S. 1988. Cobalt- and platinum-rich ferromanganese crusts and associated substrate rocks from the Marshall Islands. Marine Geology, 78(1988): 255~283.

    • Hein J R, Schultz M S, Kan J K. 1990. Insular and submarine ferromanganses minerallization of the Tonga-Lau Region. Marine Mining, 9: 305~354.

    • Hein J R, Koschinsky A. 2013a. Deep-ocean Ferromanganese Crusts and Nodules (Vol. 12). Amsterdam: Elsevier.

    • Hein J R, Mizell K, Koschinsky A, Conrad T A. 2013b. Deep-ocean mineral deposits as a source of critical metals for high- and green-technology applications: Comparison with land-based resources. Ore Geology Reviews, 51: 1~14.

    • Hesse R, Schacht U. 2011. Early Diagenesis of Deep-sea Sediments. Amsterdam: Elsevier.

    • Hey T, Tansley S, Tolle K. 2009. The fourth paradigm: Data-intensive scientific discovery. Redmond, Washington: Microsoft Research.

    • Josso P, Pelleter E, Pourret O, Fouquet Y, Etoubleau J, Cheron S, Bollinger C. 2017. A new discrimination scheme for oceanic ferromanganese deposits using high field strength and rare earth elements. Ore Geology Reviews, 87(2017): 3~15.

    • Kagesten G, Fiorentino D, Baumgartner F, Zillen L. 2019. How do continuous high-resolution models of patchy seabed habitats enhance classification schemes. Geosciences, 9: 237.

    • Kaufman L, Rousseeuw P J. 1987. Clustering by means of Medoids. Statistical Data Analysis Based on the L1-Norm and Related Methods, 405~416.

    • Kempler S, Mathews T. 2017. Earth science data analytics: Definitions, techniques and skills. International Journal of Digital Earth, 16(6): 1~8.

    • Lecours V. 2018. Habitat mapping encyclopedia of ecology. Elsevier, 89: 19~30.

    • Lehnert K, Su Y, Langmuir C H, Sarbas B, Nohl U. 2000. A global geochemical database structure for rocks. Geochemistry, Geophysics, Geosystems, 1(11): 1012.

    • Lifshits M. 2012. Lectures on Gaussian Processes. New York: Springer.

    • Li Guoqing, Liu Ying, Pang Lushen. 2018. Handbook of Big Data Analytics and Mining in Earth Science. Beijing: Posts and Telecom Press (in Chinese with English abstract).

    • Li Zhenggang, Li Huaiming, Hein J R, Dong Yanhui, Wang Mingwei, Ren Xiangwen, Wu Zhaocai, Li Xiaohu, Chu Fengyou. 2021. A possible link between seamount sector collapse and manganese nodule occurrence in the abyssal plains, NW Pacific Ocean. Ore Geology Reviews, 138: 104378.

    • Machida S, Fujinaga K, Ishii T, Nakamura K, Hirano N, Kato Y. 2016. Geology and geochemistry of ferromanganese nodules in the Japanese Exclusive Economic Zone around Minamitorishima Island. Geochemical Journal, 50: 1~17.

    • Mackay D J C. 1998. Introduction to Gaussian processes. In: Bishop C M, ed. Neural Networks and Machine Learning. Verlag: Springer, 89~93.

    • McKelvey V E, Wright N A, Bowen R W. 1983. Analysis of the world distribution of metal-rich subsea manganese nodules. Geological Survey Circular, N886.

    • Mero J L. 1965. The Mineral Resources of the Sea. Amsterdam-London-New York: Elsevier, 178~233.

    • Morgan C L. 2000. Resource Estimates of the Clarion-Clipperton Manganese Nodule Deposits. Boca Raton, Florida: CRC Press.

    • Paluszek M, Thomas S. 2019. MATLAB Machine Learning Recipes: A Problem-Solution Approach. New Jersey, USA.

    • Parsons O E. 2020. A Gaussian mixture model approach to classifying response types. In: Bouguila N, Fan W, eds. Mixture Models and Appkications, Unsupervised. Cambridge: Springer.

    • Piper D Z, Williamson M E. 1977. Composition of Pacific Ocean ferromanganese nodules. Marine Geology, 23: 285~303.

    • Price N B, Calvert S E. 1970. Compositional variation in Pacific Ocean ferromanganese nodules and its relationship to sediment accumulation rates. Marine Geology, 9(1970): 145~171.

    • Rasmussen C E, Williams C K I. 2006. Gaussian Processes for Machine Learning. Massachusetts Institute of Technology. London: MIT Press.

    • Ren Jiangbo, Deng Yinan, Lai Peixin, He Gaowen, Wang Fenlian, Yao Huiqiang, Liu Yonggang. 2021. Geochemical characteristics and genesis of the polymetallic nodules in the Pacific survey area. Earth Science Frontiers, 28(2): 412~425 (in Chinese with English abstract).

    • Ren Jiangbo, He Gaowen, Deng Xiguang, Deng Xianze, Yang Yong, Yao Huiqiang, Yang Shengxiong. 2022. Metallogenesis of Co-rich ferromaganese nodules in the northwestern Pacific: Selective enrichment of metallic elements from seawater. Ore Geology Reviews, 143(2022): 104778.

    • Trugman D T, Shearer P M. 2018. Strong correlation between stress drop and peak ground acceleration for recent M1-4 earthquakes in the San Francisco Bay area. Bulletin of the Seismological Society of America, 108: 929~945.

    • Vermeesch P. 2006. Tectonic discrimination of basalts with classification trees. Geochimica et Cosmochimica Acta, 70: 1839~1848.

    • Wang Chengshan, Hazen R M, Cheng Qiuming, Stephenson M H, Zhou Chenghu, Fox P, Sheng Shuzhong, Oberhansli R, Hou Zengqian, Ma Xiaopan, Feng Zhiqiang, Fan Junxuan, Ma Chao, Hu Xiumian, Luo Bin, Wang Juanle. 2021. The deep-time digital earth program: Data-driven discovery in geosciences. National Science Review, 8: nwab027.

    • Xu Dongyu. 2013. Ocean Mineral Geology. Beijing: Maritime Press (in Chinese with English abstract).

    • Yang Yong, He Gaowen, Ma Jinfeng, Yu Zongze, Yao Huiqiang, Deng Xiguang, Liu Fanglan, Wei Zhenquan. 2020. Acoustic quantitative analysis of ferromanganese nodule and cobalt-rich crust distribution areas using EM122 multibeam backscatter data from deep-sea basin to seamount in Western Pacific Ocean. Deep-Sea Research I, 161: 103281.

    • Yao De, Zhang Lijie, Cui Ruyong. 1996. Mineralogy and geochemistry of ferromanganese crusts from Johnston island EEZ. Marine Geology & Quaternary Geology, 16(1): 33~39 (in Chinese with English abstract).

    • Zhang Jianhui. 2007. K-means Cluster Algorithm Research and Application. Wuhan: Wuhan University of Technology Press (in Chinese with English abstract).

    • Zhang Qi, Sun Weidong, Zhao Yong, Yuan Fanglin, Jiao Shoutao, Chen Wanfeng. 2019. New discrimination diagrams for basalts based on big data research. Big Earth Data, 3(1): 45~55.

    • Zhang Tian, Ramakrishnan R, Livny M. 1996. BIRCH: An efficient data clustering databases method for very large databases. Acm Sigmod Record, 25(2): 103~114.

    • Zhao Pengda. 2019. Characteristics and rational utilization of geological big data. Earth Science Frontiers, 26(4): 1~5 (in Chinese with English abstract).

    • Zhou Yongzhang. 2018. Big Data Mining & Machine Learning in Geoscience. Guangzhou: Sun Yat-Sen University Press (in Chinese with English abstract) .

    • Zhou Yongzhang, Zuo Renguang, Liu Gang, Yuan Feng, Mao Xiancheng, Guo Yanjun, Xiao Fan, Liao Jie, Liu Yanpeng. 2021. The great-leap-forward development of mathematical geoscience during 2010-2019: Big data and artificial intelligence algorithm are changing mathematical geoscience. Bulletin of Mineralogy, Petrology, Geochemistry, 40(3): 556~573 (in Chinese with English abstract).

    • Zuo Renguang, Xiong Yihui, Wang Jian, Carranza E J M. 2019. Deep learning and its application in geochemical mapping. Earth-Science Reviews, 192: 1~14.

    • 第鹏飞, 陈万峰, 张旗, 王金荣, 汤庆艳, 焦守涛. 2018. 全球N-MORB和E-MORB分类方案对比. 岩石学报, 34(2): 264~275.

    • 冯少荣, 肖文俊. 2008. DBSCAN聚类算法的研究与改进. 中国矿业大学学报, 37(1): 105~111.

    • 李国庆, 刘莹, 庞禄申. 2018. 地球科学中的大数据分析与挖掘算法手册. 北京: 人民邮电出版社.

    • 任江波, 邓义楠, 赖佩欣, 何高文, 王汾连, 姚会强, 刘永刚. 2021. 太平洋调查区多金属结核的地球化学特征和成因. 地学前缘, 28(2): 412~425.

    • 许东禹. 2013. 大洋矿产地质学. 北京: 海洋出版社.

    • 姚德, 张丽杰, 崔汝勇. 1996. 约翰斯顿岛附近海域铁锰结核矿物学和地球化学研究. 海洋地质与第四纪地质, 16(1): 33~39.

    • 张建辉. 2007. K-means聚类算法研究及应用. 武汉: 武汉理工大学出版社.

    • 赵鹏大. 2019. 地质大数据特点及其合理开发利用. 地学前缘, 26(4): 1~5.

    • 周永章, 张良均, 张奥多, 王俊. 2018. 地球科学大数据挖掘与机器学习. 广州: 中山大学出版社.

    • 周永章, 左仁广, 刘刚, 袁峰, 毛先成, 郭艳军, 肖凡, 廖杰, 刘艳鹏. 2021. 数学地球科学跨越发展的十年: 大数据、人工智能算法正在改变地质学. 矿物岩石地球化学通报, 40(3): 556~573.