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

一直以来,探索玄武岩地球化学特征与大地构造环境之间的联系是地球化学领域的一个重要研究方向。橄榄石是岩浆最早期结晶的矿物之一,其在玄武质岩浆形成和演化过程中记录了诸多信息。鉴于此,学者们尝试利用橄榄石的元素组成判别大洋中脊玄武岩(MORB)、洋岛玄武岩(OIB)和岛弧玄武岩(IAB)三种构造环境。常用的玄武岩构造环境判别图解难以满足精度要求,于是引入机器学习算法作为判别手段来解决上述问题。机器学习判别方法的分类效果在很大程度上取决于参数选取的合理性。为此,本文提出一种耦合灰狼优化算法(GWO)和支持向量机(SVM)的智能判别方法。该方法利用GWO寻求SVM算法最优参数组合,以形成橄榄石组成元素和玄武岩构造环境之间的最佳映射关系,从而实现对MORB、OIB和IAB三种构造环境的准确判别。此外,根据公开发表的玄武岩样品的地球化学数据,结合混淆矩阵及其衍生评价指标,通过仿真实验、随机子抽样验证和k折交叉验证等方式评估了所提方法的判别性能。评估结果表明, GWO-SVM耦合判别方法在利用橄榄石成分判别玄武岩构造环境方面具有较好的分类效果,其判别准确率可达85%以上。由此可见,相较于传统判别图解方法,基于多算法融合的机器学习判别方法能够更加有效地提升构造环境判别效果。     

玄武岩类形成的大地构造环境的Th／Hf—Ta／Hf图解判别

Th,Ta,Hf是一组耐熔强亲岩浆元素，由于地球化学性质的相似性，其相互之间的比值关系能将深部作用的地球化学过程较好地恢复出来，玄武岩类，是原始玄武岩浆形成的岩石的Ta/Hf及Th/Hf比值，能较好地反映其源区的Th,Ta,Hf之间的分异特征，在一般情况下，这一特征与某种确定的大地构造环境有密切关系和确定的因果联系，可用来判别玄武岩类形成的大地构造环境及其源区Th,Ta,Hf分异特征，本文根据典型大地构造环境玄武岩类的Th,Ta,Hf数据，提出了玄武岩类形成的大地构造环境判别的Th/Hf-Ta/Hf双对数图及判别方法。     

微量元素La,Nb,Zr在判别大地构造环境方面的应用

La，Nb，Zr是一组耐熔强亲岩浆元素。其相互之间的比值关系恢复深部作用的地球化学过程。玄武岩类特别是原始玄武岩浆形成的岩石的La／Zr及Nb／Zr比值，能较好地反映其源区的La，Nb。Zr之间的分异特征，在一般情况下这一特征与某种确定的大地构造环境有密切关系，可用来判别玄武岩类形成的大地构造环境及其源区分异特征。本文根据世界上典型大地构造环境区玄武岩类的La，Nb，Zr数据。划分玄武岩形成的大地构造位置，提出了La／Zr-Nb／Zr双对数图及判别方法。     

峨眉山玄武岩系的Th、Ta、Hf特征及岩浆源区大地构造环境探讨

本文讨论了利用Th、Ta、Hf之间比值判别岩浆岩源区大地构造环境的机理,根据世界典型大地构造环境区岩浆岩系的Th、Ta、Hf数据,提出了利用Th、Ta、Hf比值判别岩浆岩源区大地构造环境的方法。在此基础上,探讨了峨眉山玄武岩系岩浆源区的大地构造环境,认为该岩系属地幔热柱成因。     

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

辉长岩是化学成分与玄武岩类似的侵入岩,前人认为它的形成过程太复杂,对应的岩浆可能经过了分离结晶作用、混染作用等,不能用Pearce判别图来判断岩浆岩形成的构造环境。本文利用GEOROC数据库的资料对辉长岩进行大数据挖掘。首先根据前人成果,将GEOROC数据库的辉长岩形成的大地构造环境分为大陆玄武岩环境、汇聚边界环境、板内火山岩环境和大洋岛弧玄武岩环境等4类;然后在数据清洗基础上,利用Python语言,依托sklearn库,实现支持向量机、K近邻和随机森林等3种机器学习算法,获得3种对应的分类器结果输出。对辉长岩的构造环境进行智能判别结果显示,随机森林方法效果最好,判断准确率可达97%,利用辉长岩的地球化学大数据来判断岩浆岩的构造环境是完全可行的。     

玄武岩类岩石大地构造环境的Th、Nb、Zr判别

本文通过研究Th、Nb、Zr的地球化学性质和判别机理，根据世界上典型大地构造环境区玄武岩类的Th、Nb、Zr数据，研究了Th、Nb、Zr判别玄武岩大地构造环境的地球化学机理，发现大同构造环境区玄武岩系的Th、Nb、Zr特征具有显著差异，其比值特征能将玄武岩形成的大地构造位置很好地划分出来，提出了判别玄武岩大地构造环境的Nb/Zr-Th/Zr双对数判别图，试图能较好地区分出大洋板块发散边缘，板块汇聚边缘，大洋板内，大陆板内及地幔热柱形成的玄武岩，得出了玄武岩大地构造环境判别的标志，指出了用Th、Nb、Zr判别玄武岩大地构造环境的方法。     

基于镁铁-超镁铁岩中单斜辉石主量元素含量的决策树集成算法对比

依靠岩浆构造环境的地球化学成分认识岩浆形成过程是岩石地球化学中的重要应用。当前利用岩石地球化学成分判别构造环境的工作还不够深入。用4种基于决策树的机器学习方法对来自全球新生代洋岛玄武岩(OIB)、岛弧玄武岩(IAB)及大洋中脊玄武岩(MORB)等镁铁-超镁铁岩中单斜辉石的13种主量元素构成数据集进行了岩浆构造环境判别和主要特征排序。通过对比4种基于决策树的机器学习方法,验证了树类算法对于地球化学成分识别问题的有效性,并总结出4种方法在处理岩浆构造环境判别问题时的优劣:决策树算法判别过程更易于理解,但是其准确率欠佳;boosting算法中的AdaBoost和GBDT对于岩浆构造环境的鉴别准确度较高,但构造过程复杂;bagging集成算法随机森林在权衡性能和模型可理解性时不失为一个良好的选择。此外,还通过4种算法的特征重要性排序得出Cr_2O_3,TFeO,TiO_2,FeO和Al_2O_3是进行岩浆构造环境判别的重要成分。     

板块汇聚边缘玄武岩大地构造环境的Th、Nb、Zr判别

根据Th、Nb、Zr的地球化学性质和判别机理，利用世界上典型大地构造环境区玄武岩类的Th、Nb、Zr数据，研究了Th、Nb、Zr判别玄武岩大地构造环境的地球化学机理，发现不同构造环境区玄武岩系的Th、Nb、Zr特征具有显著差异，其比值特征能将玄武岩形成的大地构造位置很好地划分出来。提出了板块汇聚边缘玄武岩大地构造环境的Nb/Zr Th/Zr判别图，该图不但能较好地区分出板块汇聚边缘中的洋—洋俯冲带、洋—陆俯冲带及陆—陆碰撞带，同时还能反映出岛弧区大地构造环境演化的趋势。     

藏北冈塘错地区OIB型和E-MORB型玄武岩组合及大地构造意义

通过对西藏羌塘中部冈塘错地区玄武岩地球化学特征研究,发现该区有OIB型和E-MORB型两类玄武岩.对两类玄武岩进行主量元素、稀土元素和微量元素地球化学图解分析及大地构造意义探讨,结合区域地质背景,认为冈塘错地区玄武岩形成于洋岛环境,可能受地幔柱岩浆活动影响,E-MORB型玄武岩的出现说明该区可能存在洋底高原.     

全球新生代安山岩构造环境有关问题探讨

20世纪70～80年代,以Pearce为代表的一批科学家先后提出了玄武岩和花岗岩的构造环境判别图,将构造环境与岩石地球化学特征有机地结合起来,为岩浆岩大地构造环境研究开辟了新途径。但学术界对全球广布的安山岩构造环境及相关地球化学特征问题的讨论则相对不足。本文利用GEOROC 和PetDB 两个数据库对全球新生代安山岩进行数据挖掘,讨论了它们的地球化学特征及形成环境。初步将全球新生代安山岩归属为12个形成构造环境,其中67.71%产出于岛弧、陆缘弧等汇聚板块边缘环境,其余安山岩则形成于大陆板内、大陆溢流、洋岛、大陆裂谷、洋中脊等构造环境。研究表明,常用的玄武岩微量元素判别图以及LILE/HFSE 玄武岩判别图均在一定程度上可用于安山岩成因及环境判别, 暗示安山岩地球化学成分也可用于构造环境的判定。采用大数据思维,探索洋岛安山岩（OIA）和岛弧安山岩（IAA）中地球化学元素的关联关系,从获得的近20 000 个OIA-IAA 判别图中选出lg（Cs/Ta）-lg（Cu/Ta）、lg（CaO/Nb）-lg（Cs/Zr）和lg（Cu/Ta）-lg（Co/Nb）等6个图解,能有效限定它们的构造环境,为安山岩成因及形成环境研究提供了新的思路。这些初步成果说明科学大数据的研究方法可成为岩浆岩构造环境及地球化学研究中的重要有效手段。     

An Intelligent Method for Geochemical Discrimination of Tectonic Settings of Basalt Based on Olivine Composition: GWO-SVM Method and its VerificationEI北大核心CSCD

Geochemical discrimination of tectonic settings of basalts has been an important research direction of geochemistry for decades. Olivine is one of the earliest crystallized minerals of basaltic magma, which records a lot of hidden information of the formation and evolution of the magma. Therefore, basic elements in olivine are used to discriminate three tectonic settings, including the mid-ocean ridge basalt (MORB), ocean island basalt (OIB) and island arc basalt (IAB). However, it is still difficult to accurately discriminate the tectonic settings by using these diagrams. The machine learning algorithm is introduced to solve the aforementioned problem. The classification performance of the machine learning discrimination method largely depends on the rationality of parameter determination. To this end, the paper proposes a coupling intelligent method for geochemical discrimination of tectonic settings using olivine composition of the basalts based on the grey wolf optimizer (GWO)-optimized support vector machine (SVM), or GWO-SVM for short. GWO is used to seek the optimal parameter combination of SVM to form the optimal mapping relationship between basic elements in olivine and basalt tectonic settings, so as to realize the accurate discrimination of MORB, OIB and IAB. In addition, according to the published geochemical data of basalt samples, the discrimination performance of GWO-SVM is evaluated by means of the simulation experiment, hold-out validation and k-fold cross-validation. The evaluation results are represented by the confusion matrix and its derived evaluation indicators. The results show that GWO-SVM can discriminate the tectonic settings of the basalts based on olivine compositions with overall classification accuracy of up to 85%. Thus, in comparison with the traditional discrimination diagram method, the machine learning discrimination method based on multi-algorithm fusion can significantly improve the discrimination accuracy of basalt tectonic settings. © 2020, Science Press. All right reserved.     

Geochronology and geochemistry of metasedimentary rocks from the Dongnancha Formation in the Huadian area,central Jilin Province,Northeast (NE) China: Implications for the tectonic evolution of the eastern segment of the Paleo-Asian Ocean

To constrain the tectonic evolution of the eastern segment of the Paleo-Asian Ocean, we conducted zircon U–Pb-Hf dating and whole-rock geochemical analyses for metasedimentary rocks from the Dongnancha Formation in the Huadian area in central Jilin Province, Northeastern (NE) China. Most detrital zircons from the metasedimentary rocks display clear oscillatory zoning and striped absorption in cathodoluminescence (CL) images and have Th/U ratios of 0.1–1.8, thus indicating a magmatic origin. U–Pb isotopic dating using LA-ICP-MS method for zircon samples from the metasedimentary rocks reveals that the depositional age can be constrained to the period between 250 and 222 Ma. Geochemical data reveal low to intermediate degrees of weathering of the source material and compositionally low to intermediate maturity. Detailed analyses of detrital zircon U–Pb-Hf geochronology and geochemistry show that these metasedimentary rocks are derived from a bidirectional provenance. The predominant derivation is from Permian–Early Triassic felsic-intermediate igneous rocks of central Jilin Province and adjacent regions in the northern margin of the North China Craton, although felsic-intermediate igneous rocks and continental material in the eastern segment of the Central Asian Orogenic Belt from the Cambrian–Carboniferous represent additional sources and minor amounts of Paleoproterozoic–Neoproterozoic material have been input from the North China Craton. A number of geochemical indicators and tectonic discrimination diagrams collectively indicate a continental island arc-active continental margin setting for the deposition of the protoliths of the metasedimentary rocks. The results of geochemical and geochronological analyses of the provenance and tectonic setting of the metasedimentary rocks indicate that the Dongnancha Formation was likely deposited in an intermountain basin in a post-orogenic fast uplift setting, suggesting that the final closure of the eastern segment of the Paleo-Asian Ocean in the Huadian area of central Jinlin Province likely occurred between the Early Triassic and Middle Triassic.     

机器学习在岩矿地球化学研究中的应用——综述与思考

岩石与矿物的地球化学成分数据具有高维度特征。传统的岩矿地球化学成分研究主要采用二元/三元图解判别法,准确率不高,在数理统计方法上有欠缺。机器学习方法非常适用于对大样本高维度的岩矿成分数据进行数理统计处理。本文在介绍机器学习常见算法基本原理的基础上,总结近5年来国内外学者将机器学习方法应用于岩石矿物成分数据研究的实例,包括：① 根据矿物成分溯源其母岩（源岩）、判别矿床类型,② 新生代火山岩溯源,③ 判别变质岩原岩,④ 依据岩浆岩成分判别大地构造环境等。已有的研究实例显示,机器学习方法的准确度明显优于传统的低维度判别法。机器学习本质是分析大样本数据的高维度变量之间的相关、归类等多元统计问题。推广机器学习的应用需要建设开放获取（Open Access）的矿物、岩石成分数据库,同时全面实施开放研究（Open Research）的发表策略。     

三类构造背景辉长岩单斜辉石主量元素和微量元素的数据分析研究

判别岩浆岩产出的构造环境已经成为岩石学、地球化学及其地球动力学研究的重要内容。作为岩浆岩中的一种喷出岩,玄武岩被视为判别构造环境的最佳成员。对其中单斜辉石的研究,由于其数据本身的利用程度有限而效果欠佳。理论上,不同构造环境的辉长岩也会存在一定差异。为此,利用机器学习算法研究全球新生代辉长岩的单斜辉石势在必行。本文主要针对岛弧(IAB)、洋岛(OIB)及大洋中脊(MORB)3种构造背景辉长岩的单斜辉石进行特征筛选和数据分类。从GEOROC数据库中,经数据收集与清洗,我们分别获得岛弧辉长岩单斜辉石数据385条,洋岛辉长岩单斜辉石数据756条,大洋中脊辉长岩单斜辉石数据5 500条。其中绝大部分为主量元素数据,其余为微量元素数据。在特征提取部分,我们选用卡方检验判断特征独立性,F检验估计两个随机变量之间的线性依赖程度,互信息法捕获其他种类的统计相关性。3种检验方法互相印证,得出了统计学可靠的重要分类特征。在数据分类过程中,本文对比了K-近邻、决策树和支持向量机3种主流机器学习分类算法在辉长岩数据上的表现。研究表明,对于上述3种构造背景,Al2O3、TiO2为最有区分度的辉长岩单斜辉石主量元素成分,Sr为最有区分度的微量元素成分。另外,对于3种构造背景的辉长岩单斜辉石主量元素和微量元素数据,机器学习模型分类准确率均达94%。     

内蒙古北山地区小红山钒钛磁铁矿区双峰式侵入岩——岩石学、年代学、地球化学特征及其地质意义

对内蒙古北山地区小红山钒钛磁铁矿区内侵入岩的岩石学、锆石U-Pb年代学和全岩地球化学资研究显示,该岩体岩性为辉长岩和花岗岩,LA-ICP-MS锆石U-Pb测年结果表明成岩年龄分别为431.1±2.4 Ma和424.9±2.3 Ma,形成时代相近,属同一构造岩浆作用事件. 该套侵入岩的SiO₂含量呈双峰式,稀土、微量元素特征表明它们具有岛弧岩浆岩地球化学特征,其源区可能受到了俯冲流体交代作用的影响. 地球化学特征指示小红山花岗岩源于下地壳物质在高温条件下的部分熔融,辉长岩为富集地幔部分熔融的产物. 结合区域地质背景及构造判别,认为小红山中晚志留世双峰式岩浆组合是北山洋南向俯冲诱导大陆边缘伸展环境下的产物.     

藏南冈底斯带中段始新世岩浆作用的厘定及其大地构造意义

本文对藏南冈底斯带中段的花岗岩类和角闪辉长岩进行了锆石U-Pb年代学和全岩地球化学分析,据此阐明了岩体的形成机制与演化过程,并探讨了成岩时的大地构造背景。分析结果显示,研究区内花岗岩类和角闪辉长岩体的LA-ICPMS锆石U-Pb定年结果为41~55Ma,为始新世早-中期岩浆活动的产物,代表了区内岩体的成岩年龄。在地球化学组成上,花岗岩类属于钙碱性到高钾钙碱性系列,均富集轻稀土(LREE)和大离子亲石元素(LILE)(Rb、Ba和K),强烈亏损Nb、Ta、P等高场强元素(HFSE),具有弧型岩浆岩的地球化学组成。此外,花岗岩类的铝饱和指数(A/CNK)小于1.1,属于准铝质到弱过铝质的I型花岗岩。角闪辉长岩为石榴橄榄岩部分熔融的产物,并在后期侵位的过程中遭受到了壳源物质的混染。综合分析表明,研究区内的岩体形成于初始碰撞向主碰撞的转化阶段。始新世早期(～50Ma)新特提斯洋板片的断离引起软流圈物质上涌,导致岩石圈地幔发生部分熔融形成基性岩浆,随后基性岩浆底侵至下地壳并诱发下地壳发生部分熔融形成花岗岩质岩浆,最后经过岩浆混合作用形成始新世早-中期冈底斯地区的花岗岩类。     

Tectonic discrimination of basalts with classification trees

Traditionally, geochemical classification of basaltic rocks of unknown tectonic affinity has been performed by discrimination diagrams. Although easy to use, this method is fairly inaccurate because it only uses bi- or trivariate data. Furthermore, many popular discrimination diagrams are statistically not very rigorous because the decision boundaries are drawn by eye, and they ignore closure, thus violating the rules of compositional data analysis. Classification trees approximate the data space by a stepwise constant function, and are a more rigorous and potentially more effective way to determine tectonic affinity. Trees allow the simultaneous use of an unlimited number of geochemical features, while still permitting visualization by an easy-to-use, two-dimensional graph. Two classification trees are presented for the discrimination of basalts of mid-ocean ridge, ocean island, and island arc affinities. The first tree uses 51 major, minor, and trace elements and isotopic ratios and should be used for the classification of fresh basalt samples. A second tree only uses high field strength element analyses and isotopic ratios, and can also be used for basalts that have undergone alteration. The probability of successful classification is 89% for the first and 84% for the second tree, as determined by 10-fold cross-validation. Even though the trees presented in this paper use many geochemical features, it is not a problem if some of these are missing in the unknown sample. Classification trees solve this problem with surrogate variables, which give more or less the same decision as the primary variables. The advantages of the classification tree approach over discrimination diagrams are illustrated by a comparative test on a sample dataset of known tectonic affinities. Although arguably better than discrimination diagrams, classification trees are not perfect, and the limitations of the method are illustrated on a published dataset of basalts from the Pindos Basin (Greece).     

Petrogenesis and tectonic implications of the late Carboniferous calc-alkaline and shoshonitic magmatic rocks in the Awulale mountain,western Tianshan

The widespread late Carboniferous calc-alkaline and shoshonitic magmatic rocks in the Awulale mountain provide crucial constraints on the tectonic evolution of the western Tianshan. Here, we perform detailed petrological investigations as well as zircon U-Pb chronological, whole-rock geochemical and Sr-Nd isotopic analyses on these magmatic rocks from two geological sections along the Duku road. Magmatic rocks in the section I with zircon SHRIMP U-Pb ages of 306.8 Ma and 306.4 Ma are composed of medium-K calc-alkaline to shoshonitic basalt, trachy-andesite and trachyte, while those in the section II consist of shoshonitic trachy-andesite, trachyte with a U-Pb age of 308.1 Ma, and monzonite with a U-Pb age of 309.6 Ma. All these magmatic rocks are characterized by strong enrichments in large iron lithophile elements with depletions of Nb, Ta and Ti, indicating the origination from subduction-modified lithospheric mantle. The ε_Nd(t) values of the rock samples collected from the section I (2.80–5.45) and section II (3.34–5.37) are generally higher than those of the Devonian to early Carboniferous arc-type magmatic rocks in the Yili-central Tianshan, suggesting that depleted asthenosphere might also be involved in their generation. Based on these geochemical data and petrological observations, we suggest that the early-stage (308.1–309.6 Ma) shoshonitic monzonite, trachy-andesite and trachyte in the section II were generated by mixing between mafic magmas and trachytic melts, while the late-stage (306.4–306.8 Ma) medium-K calc-alkaline to shoshonitic basalt, trachy-andesite and trachyte in the section I were produced by partial melting of depleted asthenospheric and metasomatized lithospheric mantle, followed by the processes of fractional crystallization and crustal contamination. Taking into account the available regional geological data, the subduction of south Tianshan ocean was probably ceased at ∼310 Ma, and these calc-alkaline and shoshonitic magmatic rocks in the Awulale mountain formed in a post-collisional setting subsequent to slab break-off.     

Petrology,zircon U–Pb geochronology and tectonic implications of the A1-type intrusions: Keban region,eastern Turkey

The Southeast Anatolian Orogenic Belt (SAOB), the most extensive segment of the Alpine-Himalayan Orogenic Belt, resulted from the northward subduction of the southern branch of the Neotethys oceanic crust beneath the Anatolian micro-plate. We present new whole-rock geochemistry, zircon U–Pb ages, and Lu–Hf isotope data from the stocks and dykes with a length of up to tens of meters belonging to the Keban magmatic rocks, eastern Turkey. These rocks are represented by syenite and quartz monzonite intruded into the Keban metamorphic complex. The geochemistry data indicates that the samples bear mostly metaluminous, variably high alkalines (K₂O + Na₂O), Ga/Al ratios and zircon saturation temperature, and typically the A-type granite characters. According to the Y/Nb vs Yb/Ta diagram, the Keban magmatic rocks show A₁-type geochemical signatures modified by crustal melts. Syenite and quartz monzonite samples from Keban magmatic rocks give zircon U–Pb ages of 77.4 ± 0.34 Ma, 76.3 ± 0.3 Ma and 76.36 ± 0.34 Ma, respectively. On the primitive mantle-normalised trace element patterns, the Keban magmatic rocks show enrichment in large-ion lithophile elements (LILEs) relative to high field strength elements (HFSEs). They are coupled with slightly negative Nb–Ta anomalies. Chondrite-normalised rare earth-element patterns show strong enrichment in LREEs relative to HREEs, a typical A-type granites feature. The zircons have negative εHf_(t) values that vary from ?2.68 to ?0.41, and Hf model ages (T_DM2) range from 1171.54 to 1329.26 Ma, indicating the enriched lithospheric mantle sources and crustal contribution. The sources and evolution of the alkaline magmas might be related to the post-collisional tectonic setting.