Quantitative prediction of molybdenum-copper polymetallic mineral resources in the Xindalai grassland-covered area of Inner Mongolia based on geological anomalies

doi:10.13745/j.esf.sf.2021.1.16

Abstract

Abstract:

The Xindalai grassland-covered area of Inner Mongolia, in the western part of the Erlian-Dongwuqi molybdenum-copper polymetallic belt of the Paleo-Asian metallogenic domain, distributes endogenous granophile metal deposits with favorable ore-forming conditions, such as the Wulandele copper-molybdenum deposit and the Zhunsujihua molybdenum deposit. However, due to the influence of large herbage and Quaternary overlays, all kinds of mineralization/ore indicators in this area are indirect, mixed, concealed, weak or incomplete, causing considerable uncertainty and hazard to ore prospecting and exploration. Therefore, it is necessary to develop a quantitative prediction model to guide ore prospecting in the overlay area. In this paper, guided by the geological anomaly theory developed originally by Zhao et al., we analyzed the diversity of mineralization in Xindalai and its adjacent areas and summarized the vertical distribution of mineralization data. Using the S-A multifractal filtering model, the overlay interference on soil geochemical and high-precision magnetic survey data is reduced, and weak anomalies from deep sources are identified and extracted. The extracted information on faults, Jurassic granitic rocks, dykes and rock mass-wall rock contact zones, as well as on PC1-PC2 element combination anomalies, high-precision geomagnetic anomalies, and geographic location (X-Y coordinates) of metallogenic and non-metallogenic units, were taken as input variables using the random forest (RF) prediction method. The SMOTE sampling technology was used to overcome training sample insufficiency caused by limited number of ore deposits/occurrences in the grassland-covered area. Eventually the comprehensive geo-anomalies closely related to mineralization were quantified after five hundred iterations in the RF simulation. The simulation results show that the average OOB error and ACU value were 2.26% and 0.972, respectively, and 88.46% of known ore deposits/occurrences correspond to geo-anomalies with metallogenic advantage ≥0.783, demonstrating the effectiveness of the prediction method. In order to further reduce the exploration risk, we performed a risk-return analysis to show that 25 out of 26 ore deposits/occurrences were distributed in the positive return range, and only 3 ore occurrences were associated with medium to high risk values. On this basis, we used the metallogenic advantages of geo-anomalies associated with low-risk, high-return areas to re-map the grassland-covered area and ultimately delineated the preferable ore-finding district.

Key words: geo-anomaly, quantitative prediction of mineral resources, risk-return analysis, Xindalai grassland-covered area

CLC Number:

P612
P628

XIA Qinglin, ZHAO Mengyu, WANG Xiaochen, LENG Shuai, LI Tongfei, XIONG Shuangcai. Quantitative prediction of molybdenum-copper polymetallic mineral resources in the Xindalai grassland-covered area of Inner Mongolia based on geological anomalies[J]. Earth Science Frontiers, 2021, 28(3): 56-66.

Figures/Tables 13

Fig.1 Simplified metallogenic map of the Xindalai grassland-covered area. Modified after [4].

Fig.2 Vertical distribution of orebodies in the Xindalai grassland-covered area

Fig.3 The prediction error curves with different over-sampling ratios

Fig.4 Distribution of buffer zones for different geological features

Fig.5 The S-A anomaly distribution map of main geochemical elements

Fig.6 The S-A anomaly distribution map of high-precision geomagnetic data

Fig.7 The relation curves between prediction error and decision tree number for different testing conditions

Tab1e 1 RF model performance parameters

观测	预测		分类错误率/%
观测	不含矿	含矿	分类错误率/%
不含矿	156	8	4.88
含矿	1	233	0.43
OOB error	2.26%
AUC	0.972

Fig.8 ROC curve for mineral resources prediction

Fig.9 Extracted comprehensive geo-anomalies using RF method based on the SMOTE dataset

Fig.10 Results of return (a) and risk (b) analysis

Fig.11 Risk vs. return scatter plot

Fig.12 Distribution of preferable molybdenum-copper polymetallic mineral ore-finding areas based on geo-anomalies

References 28

[1]	赵鹏大, 池顺都. 初论地质异常[J]. 地球科学: 中国地质大学学报, 1991, 16(3):241-248.
[2]	赵鹏大, 陈永清, 刘吉平, 等. 地质异常成矿预测理论与实践[M]. 武汉: 中国地质大学出版社, 1999.
[3]	ZHAO P D, XIA Q L. “5P” areas: a new approach for delineation targets in mineral prediction[C]. Фундаментальные проблемы геологии месторождений полезных ископаемых и металлогении, 2010, 1:234-243.
[4]	熊双才, 夏庆霖, 姚春亮, 等. 内蒙古新达来草原覆盖区钼铜多金属成矿预测[J]. 地质学刊, 2015, 39(3):422-430.
[5]	邵积东, 王守光, 赵文涛, 等. 内蒙古中部地区重要成矿区带成矿地质特征及找矿潜力分析[J]. 西部资源, 2008(1):59-61.
[6]	任纪舜, 陈廷愚, 牛宝贵. 中国东部及邻区大陆岩石圈的构造演化与成矿[M]. 北京: 科学出版社, 1990.
[7]	XIAO W J, WINDLEY B F, HAO J, et al. Accretion leading to collision and the Permian Solonker suture, Inner Mongolia, China: termination of the central Asian orogenic belt[J]. Tectonic, 2003, 22(6):1069-1088.
[8]	赵鹏大. “三联式”资源定量预测与评价: 数字找矿理论与实践探讨[J]. 地球科学: 中国地质大学学报, 2002, 27(5):482-489.
[9]	刘翼飞, 聂凤军, 江思宏, 等. 内蒙古苏尼特左旗准苏吉花钼矿床成岩成矿年代学及其地质意义[J]. 矿床地质, 2012, 31(1):119-128.
[10]	成秋明, 夏庆霖. 覆盖区矿产综合预测思路与实践[J]. 矿物学报, 2011(增刊):755-756.
[11]	成秋明. 覆盖区矿产综合预测思路与方法[J]. 地球科学, 2012, 37(6):1109-1125.
[12]	BREIMAN L. Random forest[J]. Machine Learning, 2001, 45:5-32. DOI URL
[13]	BREIMAN L. Bagging predictors[J]. Machine Learning, 1996, 24(2):123-140.
[14]	BREIMAN L, FRIEDMAN J H, OLSHEN R A, et al. Classification and Regression trees[J]. Biometrics, 1984, 40(3):358.
[15]	陈进, 毛先成, 刘占坤, 等. 基于随机森林算法的大尹格庄金矿床三维成矿预测[J]. 大地构造与成矿学, 2020, 44(2):231-241.
[16]	柯元楚, 史忠奎, 李培军, 等. 基于Hyperion高光谱数据和随机森林方法的岩性分类与分析[J]. 岩石学报, 2018, 34(7):2181-2188.
[17]	FORBES A D. Classification-algorithm evaluation: five performance measures based on confusion matrices[J]. Journal of Clinical Monitoring, 1995, 11(3):189-206. DOI URL
[18]	CARLETTA J. Assessing agreement on classification tasks: the kappa statistic[J]. Computational Linguistics, 1996, 22(2):249-254.
[19]	METZ C E. Basic principles of ROC analysis[J]. Seminars in Nuclear Medicine, 1978, 8(4):283-298. DOI URL
[20]	WANG Z Y, YIN Z, CAERS J, et al. A Monte Carlo-based framework for risk-return analysis in mineral prospectivity mapping[J]. Geoscience Frontiers, 2020, 11(6):2297-2308. DOI URL
[21]	CARRANZA E J M, LABORTE A G. Random forest predictive modeling of mineral prospectivity with small number of prospects and data with missing values in Abra (Philippines)[J]. Computers & Geosciences, 2015, 74:60-70. DOI URL
[22]	CHAWLA N V, BOWYER K W, HALL L O, et al. SMOTE: synthetic minority over-sampling technique[J]. Journal of Artificial Intelligence Research, 2002, 16(1):321-357. DOI URL
[23]	HARIHARAN S, TIRODKAR S, PORWAL A, et al. Random forest-based prospectivity modelling of Greenfield Terrains using sparse deposit data: an example from the Tanami Region, Western Australia[J]. Natural Resources Research, 2017, 26(4):489-507. DOI URL
[24]	LI T, XIA Q, ZHAO M, et al. Prospectivity mapping for Tungsten polymetallic mineral resources, Nanling Metallogenic Belt, South China: use of random forest algorithm from a perspective of data imbalance[J]. Natural Resources Research, 2020, 29(1):203-227. DOI URL
[25]	FILZMOSER P, HRON K, REIMANN C. Principal component analysis for compositional data with outliers[J]. Environmetrics, 2010, 20(6):621-632. DOI URL
[26]	CHENG Q M, XU Y G, GRUNSKY E. Integrated spatial and spectrum method for geochemical anomaly separation[J]. Natural Resources Research, 2000, 9(1):43-52. DOI URL
[27]	CHENG Q M, AGTERBERG F P, BALLANTYNE S B. The separation of geochemical anomalies from background by fractal methods[J]. Journal of Geochemical Exploration, 1994, 51(2):109-130. DOI URL
[28]	ZUO R G, WANG Z Y. Effects of random negative training samples on mineral prospectivity mapping[J]. Natural Resources Research, 2020, 29(6):3443-3455. DOI URL