深时铁矿物分布特征及演化趋势预测初探

doi:10.19657/j.geoscience.1000-8527.2025.040

摘要/Abstract

摘要：

矿物是数十亿年地球系统演化的重要信息载体,以数据驱动的矿物学研究对于深入挖掘地球演化过程中的潜在演替规律和驱动机制具有重大意义。铁元素(Fe)在地球物质和能量循环、生命演化、环境修复等方面具有重要作用,也是一种常见的指示环境变迁的氧化还原敏感元素。本文通过探讨949种铁矿物多样性演化规律及分布特征,初步构建起深时铁矿物演化与地球环境演变及生命演化之间的内在联系。研究结果表明,深时铁矿物多样性呈幕式增长特征,增长高峰期与超大陆碰撞期相吻合,板块运动、大气增氧、生命代谢活动等过程共同促进铁矿物向更复杂、更多元的方向演化。我们进一步利用BP神经网络(BPNN)、随机森林(RF)和支持向量机回归(SVR)3种机器学习算法建立铁矿物演化趋势预测模型,其中RF相较于BP神经网络和支持向量机回归预测精度更高、泛化能力更强。

关键词: 铁矿物, 矿物演化, 机器学习, 演化特征

Abstract:

Minerals are important recorders of Earth’s evolution over billions of years. Data-driven research in mineralogy makes it easier to understand the underlying laws and driving mechanisms of the evolutionary processes of Earth. Iron plays a critical role in the matter and energy cycles, evolution of life, and environmental remediation, and is also recognized as a vitally important ‘oxygen fugacity buffer’ during the evolution of Earth surficial system. A total amount of 949 iron minerals are collected in this study. And by exploring their evolutionary patterns and distributional characteristics, a preliminary framework is constructed to tap into the intrinsic links between the evolution of deep-time iron minerals, evolution of Earth’s environment and evolution of life. The results reveal that the diversity of iron minerals has expanded episodically throughout geological history, and the peak of the growth stage coincides with the period of supercontinent collisions. The processes of plate motion, atmospheric oxygenation, and life’s metabolic activities have combined to promote the evolution of iron minerals in a more complex and diverse direction. Furthermore, back propagation neural network (BPNN), random forest (RF) and support vector regression (SVR), are used to establish and predict the evolution model of iron minerals since 4.0 Ga. The results show that RF is more appropriate to predict the evolution trend more accurately with stronger generalization ability than BPNN and SVR.

Key words: iron mineral, mineral evolution, machine learning, evolutionary feature

中图分类号:

黄思艺, 徐靖博, 魏林宏, 蔡元峰. 深时铁矿物分布特征及演化趋势预测初探[J]. 现代地质, 2025, 39(03): 552-559.

HUANG Siyi, XU Jingbo, WEI Linhong, CAI Yuanfeng. A Preliminary Study on the Temporal Distribution Characteristics of Iron Minerals and Prediction of the Deep-Time Evolution[J]. Geoscience, 2025, 39(03): 552-559.

图/表 5

表1 铁矿物演化趋势预测模型输入参数

Table 1 Input parameters of the model for predicting the evolution trend of iron minerals

参数	单位	最大值	最小值	参考文献
大气O₂含量[lg(pO₂)]	PAL	-0.1811	-7.3256	[15]
大气CO₂含量	bar	0.2917	0.0003	[16]
海洋pH		8.20	6.64	[16]
地表温度	K	292.18	283.97	[16]
深海温度	K	279.81	271.14	[16]

表2 三种机器学习模型最优参数组合

Table 2 Optimal parameter structure for BPNN, RF, SVR.

模型	参数	最优值
BPNN	激活函数	Sigmoid
	隐藏层	1
	隐藏层神经元节点数	12
RF	决策树数目	246
	决策树最大深度	50
	最小叶子节点样本数	4
	内部节点再划分所需最小样本数	20
	分裂考虑最大特征数	3
SVR	核函数	RBF
	惩罚因子C	223.26
	核函数系数	0.49

图1 地球中铁矿物、大气氧浓度(pO2)及生命的深时共演化模式图 (a)40亿年以来铁矿物价态和种类演化散点图;(b)不同价态(Fe0、Fe2+、Fe3+、Fe2+ & Fe3+)铁矿物种类数随时间演化趋势图,pO2变化观点用矩形和箭头表示,粉红色部分代表大气氧含量变化,箭头代表可能存在的局域短暂增氧事件 [21];(c)古元古代至早寒武世真核生物类群的堆叠物种多样性 [23],古生代(约540~240 Ma)海洋生物多样性曲线 [24],上方蓝色图形代表生命演化进程 [15];浅灰色长方形代表五个超级大陆碰撞期 [25],浅紫色长方形代表GOE(大氧化事件)和NOE(新元古代氧化事件)的大致时间区间 [21]

Fig.1 Co-evolution of iron mineral diversities,pO2, and life over the last 4 billion years on the earth

图2 BPNN、RF、SVR三种机器学习模型预测铁矿物演化趋势性能对比图

Fig.2 Comparison of the performance of BPNN, RF, SVR, in predicting the evolution trend of iron minerals

表3 三种机器学习模型预测深时铁矿物演化趋势性能对比

Table 3 Predictive performance comparison of BPNN, RF, SVR models

模型	BPNN	RF	SVR
R²	0.99904	0.99904	0.99898
RMSE	7.3719	7.3603	7.6073
MAPE	2.4136	1.6361	7.3626

参考文献 44

[1]	郭承基, 王中刚. 矿物演化[J]. 矿物学报, 1981, 1(1): 1-9.
[2]	陈光远. 成因矿物学与找矿矿物学发展现状[J]. 矿物岩石地球化学通报, 1987, 6(2): 68-70.
[3]	HAZEN R M, PAPINEAU D, BLEEKER W, et al. Mineral evolution[J]. American Mineralogist, 2008, 93(11/12): 1693-1720.
[4]	HAZEN R M, MORRISON S M. On the paragenetic modes of minerals: A mineral evolution perspective[J]. American Mineralogist, 2022, 107(7): 1262-1287.
[5]	KENDALL B, ANBAR A D, KAPPLER A, et al. The global iron cycle[M]. KNOLL A H, CANFIELD D E, KONHAUSER K O. Fundamentals of geobiology. America: Blackwell Publishing Ltd. 2012: 65-92.
[6]	WADE J, BYRNE D J, BALLENTINE C J, et al. Temporal variation of planetary iron as a driver of evolution[J]. Proceedings of the National Academy of Sciences of the United States of America, 2021, 118(51): e2109865118.
[7]	KAPPLER A, BRYCE C, MANSOR M, et al. An evolving view on biogeochemical cycling of iron[J]. Nature Reviews Microbiology, 2021, 19(6): 360-374. DOI PMID
[8]	吴琪, 仇巍巍, 江拓, 等. 中国铁矿勘探开发及供需形势[J]. 地质论评, 2024, 70(A1): 111-112.
[9]	WANG P, FU F G, LIU T G. A review of the new multifunctional nano zero-valent iron composites for wastewater treatment: Emergence, preparation, optimization and mechanism[J]. Chemosphere, 2021, 285(0): 131435.
[10]	GONG Y S, WANG Y, LIN N P, et al. Iron-based materials for simultaneous removal of heavy metal(loid)s and emerging organic contaminants from the aquatic environment: Recent advances and perspectives[J]. Environmental Pollution, 2022, 299: 118871.
[11]	HAZEN R M, DOWNS R T, ELEISH A, et al. Data-driven discovery in mineralogy: Recent advances in data resources, analysis, and visualization[J]. Engineering, 2019, 5(3): 397-405.
[12]	LU A H. Mineral evolution heralds a new era for mineralogy[J]. American Mineralogist, 2022, 107(7): 1217-1218.
[13]	WANG C S, HAZEN R M, CHENG Q M, et al. The deep-time digital earth program: data-driven discovery in geosciences[J]. National Science Review, 2021, 8(9): 156-166.
[14]	李晓彦, 张超. 大数据和机器学习在矿物学研究中的应用[J]. 矿物岩石地球化学通报, 2023, 42(2): 253-266, 252.
[15]	CHEN G X, CHENG Q M, LYONS T W, et al. Reconstructing Earth’s atmospheric oxygenation history using machine learning[J]. Nature Communications, 2022, 13(1): 5862.
[16]	KRISSANSEN-TOTTON J, ARNEY G N, CATLING D C. Constraining the climate and ocean pH of the early Earth with a geological carbon cycle model[J]. Proceedings of the National Academy of Sciences of the United States of America, 2018, 115(16): 4105-4110.
[17]	LI J, CHENG J H, SHI J Y, et al. Brief introduction of back propagation (BP) neural network algorithm and its improvement[M]. DAVID J, SALLY L. Advances in computer science and information engineering: Volume 2. 2012: 553-558.
[18]	沈花玉, 王兆霞, 高成耀, 等. BP神经网络隐含层单元数的确定[J]. 天津理工大学学报, 2008, 24 (5): 13-15.
[19]	BREIMAN L. Random forests[J]. Machine Learning, 2001, 45(1): 5-32.
[20]	KHADEMIAN M, IMLAY J A. How microbes evolved to tolerate oxygen[J]. Trends in Microbiology, 2021, 29(5): 428-440. DOI PMID
[21]	LYONS T W, DIAMOND C W, PLANAVSKY N J, et al. Oxygenation, life, and the planetary system during Earth’s middle history: An overview[J]. Astrobiology, 2021, 21(8): 906-923.
[22]	HAZEN R M. Paleomineralogy of the Hadean Eon: A preliminary species list[J]. American Journal of Science, 2013, 313(9): 807-843.
[23]	TANG Q, ZHENG W T, ZHANG S H, et al. Quantifying the global biodiversity of Proterozoic eukaryotes[J]. Science, 2024, 386: eadm9137.
[24]	FAN J X, SHEN S Z, ERWIN D H, et al. A high-resolution summary of Cambrian to Early Triassic marine invertebrate biodiversity[J]. Science, 2020, 367(6475): 272-277.
[25]	HAZEN R M, LIU X-M, DOWNS R T, et al. Mineral evolution: Episodic metallogenesis, the supercontinent cycle, and the coevolving geosphere and biosphere[M]. KELLEY K D, GOLDEN H C. Building exploration capability for the 21st century. Littleton: Society of Economic Geologists. 2014: 1-15.
[26]	翟明国. 华北前寒武纪成矿系统与重大地质事件的联系[J]. 岩石学报, 2013, 29(5): 1759-1773.
[27]	BEKKER A, SLACK J F, PLANAVSKY N, et al. Iron formation: The sedimentary product of a complex interplay among mantle, tectonic, oceanic, and biospheric processes[J]. Economic Geology, 2010, 105(3): 467-508.
[28]	KONHAUSER K O, PLANAVSKY N J, HARDISTY D S, et al. Iron formations: A global record of Neoarchaean to Palaeoproterozoic environmental history[J]. Earth-Science Reviews, 2017, 172(1): 140-177.
[29]	KLEIN C. Some Precambrian banded iron-formations (BIFs) from around the world: Their age, geologic setting, mineralogy, metamorphism, geochemistry, and origins[J]. American Mineralogist, 2005, 90(10): 1473-1499.
[30]	SCHWERTMANN U, MURAD E. Effect of pH on the formation of goethite and hematite from ferrihydrite[J]. Clays and Clay Minerals, 1983, 31(4): 277-284.
[31]	OHMOTO H. Nonredox transformations of magnetite-hematite in hydrothermal systems[J]. Economic Geology, 2003, 98(1): 157-161.
[32]	宋颜, 董少春, 胡欢, 等. 基于大数据的铌钽矿物全球时空分布特征分析[J]. 地学前缘, 2023, 30(5): 197-204. DOI
[33]	唐鸣昊, 滕辉, 陆现彩, 等. 砷矿物演化与矿物生态初探[J]. 矿物岩石地球化学通报, 2024, 43(2): 418-427.
[34]	ZHUANG Z Y, ZHANG Y, LI Y, et al. Evolutionary dynamics of redox-sensitive minerals reveal details and possible regulatory mechanisms of Earth’s oxygenation events[J]. Earth and Planetary Science Letters, 2024, 626: 118528.
[35]	TANG H S, CHEN Y J. Global glaciations and atmospheric change at ca. 2.3 Ga[J]. Geoscience Frontiers, 2013, 4(5): 583-596.
[36]	HOFFMANN K H, CONDON D J, BOWRING S A, et al. U-Pb zircon date from the Neoproterozoic Ghaub Formation, Namibia: Constraints on Marinoan Glaciation[J]. Geology, 2004, 32(9): 817-820.
[37]	ZHANG S H, JIANG G Q, HAN Y G. The age of the Nantuo Formation and Nantuo glaciation in South China[J]. TERRA NOVA, 2008, 20(4): 289-294.
[38]	MACDONALD F A, SCHMITZ M D, CROWLEY J L, et al. Calibrating the Cryogenian[J]. Science, 2010, 327: 1241-1243. DOI PMID
[39]	赵彦彦, 郑永飞. 全球新元古代冰期的记录和时限[J]. 岩石学报, 2011, 27(2): 545-565.
[40]	MIAO L Y, YIN Z J, KNOLL A H, et al. 1.63-billion-year-old multicellular eukaryotes from the Chuanlinggou Formation in North China[J]. Science advances, 2024, 10(4): eadk3208.
[41]	ILBERT M, BONNEFOY V. Insight into the evolution of the iron oxidation pathways[J]. Biochimica et Biophysica Acta-Bioenergetics, 2013, 1827(2): 161-175.
[42]	王芙仙, 郑世玲, 邱浩, 等. 铁还原细菌Shewanella oneidensis MR-4诱导水合氧化铁形成蓝铁矿的过程[J]. 微生物学报, 2018, 58(4): 573-583.
[43]	张志飞, 刘璠, 梁悦, 等. 寒武纪生命大爆发与地球生态系统起源演化[J]. 西北大学学报(自然科学版), 2021, 51(6): 1065-1106.
[44]	朱永官, 段桂兰, 陈保冬, 等. 土壤-微生物-植物系统中矿物风化与元素循环[J]. 中国科学(地球科学), 2014, 44(6): 1107-1116.