图1 矿产资源定量预测学科发展历史

Fig.1 Historical development of the discipline of quantitative mineral resource prediction

图2 覆盖区找矿或深部矿产预测的诸多困难示意图:弱信息提取(左图);复杂叠加混合数据与成矿背景与异常分解(中图);不完善或缺失多元数据成矿信息的综合与集成(右图)

Fig.2 Schematic diagram illustrating key challenges in mineral exploration in covered regions or deep mineral resource prediction:Weak anomaly extraction (left); Decomposition of complexly mixed and superimposed data into backgrounds and anomalies (middle); Integration and synthesis of incomplete or missing multi-source metallogenic information (right)

图3 非线性矿产勘查理论与方法体系

Fig.3 Framework of nonlinear theories and methods in mineral resource prediction

图4 通过地球化学元素与重磁异常综合模型预测出小大青山—红娘山钼多金属“半隐伏”成矿带。识别出5个等距分布的中酸性侵入岩体(含隐伏体)和潜在矿化集中区,包含曹四夭特大型钼矿和泉子沟中型钼矿以及未知预测靶区。

Fig.4 An integrated model combining geochemical and gravity-magnetic anomalies predicts a “semi-concealed” Mo-polymetallic metallogenic belt in the Xiaodaqingshan-Hongniangshan area. Five evenly spaced intermediate-acid intrusive bodies (including concealed ones) and potential mineralization concentration zones are delineated, encompassing the Caosiyao giant Mo deposit, Quanzigou medium-sized Mo deposit, and undiscovered prospective targets.

图5 深部矿产资源预测评价相关的“科学问题、关键技术以及空间定位目标”

Fig.5 Critical scientific questions, key technologies, and targeting objectives related to deep mineral resource prediction and assessment

图6 研究区与构造背景、成矿作用、矿集区分布 (上)与典型矿床类型和成矿大地构造背景剖面示意图(下)

Fig.6 Study areas showing tectonic settings, mineralization processes, and ore district distribution (Upper) and schematic cross-section of typical deposit types and metallogenic geotectonic background (Lower)

图7 研究内容之间的逻辑关系图

Fig.7 Framework showing relationship diagram among research contents

图8 奇异性、广义自相似、分形谱系原理和模型示意图。奇异性指数Δα用于深部源引起的弱信息提取(左图); 成矿结构的广义自相似性用于深部矿体内插或外推预测(中图); 成矿系列和矿床谱系结构随深度变化规律,用于深部矿预测(右图)。

Fig.8 Schematic diagram illustrating the principles and models of singularity, generalized self-similarity, and fractal spectra. The singularity index Δα used to extract weak signals caused by deep sources (left); the generalized self-similarity of mineralization structures supports interpolation or extrapolation of deep ore bodies (middle); the variation patterns of mineralization series and ore deposit spectra with depth as a basis for deep mineral prediction (right).

图9 金川铜镍硫化物矿床自相似结构模型构建。(A)—金川矿区构造-岩体分布简图;(B)—逆冲推覆构造导致岩体-构造呈现三维自相似模式;(C)—Ⅱ矿区50行地质剖面图。其显示已知矿体下端呈现尖灭再现趋势,尖灭端新发现新矿体,并有向深部逐渐增厚的趋势。

Fig.9 Construction of a self-similar structural model for the Jinchuan Cu-Ni sulfide deposit. (A) Simplified map showing the tectonic and magmatic rock distribution in the Jinchuan mining area. (B) The thrusting and overthrusting tectonics result in a 3D self-similar pattern of intrusion-structure relationships. (C) A geological cross-section showing a known ore body exhibiting a pinch-and-swell pattern; at the pinch-out end, a new ore body was discovered.

图10 沿环太平洋和特提斯成矿域中新生代斑岩铜矿床分布图(数据来源于USGS data)

Fig.10 Distribution map of Mesozoic to Cenozoic porphyry copper deposits along the Circum-Pacific and Tethyan metallogenic belts (Data source: USGS)

图11 数学地球科学学科知识体系与矿产预测模型构建知识库

Fig.11 Knowledge system of Mathematical Geosciences and its integration into the knowledge base for mineral prediction model development

图12 斑岩铜矿知识图谱和分析系统

Fig.12 Knowledge graph and analytical system for porphyry copper deposits

图13 斑岩铜矿知识图谱查询结果显示Cu和S元素在成矿地质过程的关联性。圆形符号表示节点(地质术语), 箭头连接线表示关联性,圆型的大小表示出现频率。

Fig.13 Query results from the knowledge graph of porphyry copper deposits, illustrating the relationship between Cu and S elements in the mineralization process. Circular symbols represent nodes (geological terms), the arrowed links indicate relationships, and the size of a circle represents the frequency of the node’s occurrence.

图14 矿产资源智能预测平台体系构架:多功能智能体关联、智能预测平台、数据库与计算资源库、知识图谱和大语言、预测计算大模型

Fig.14 Architecture of the intelligent mineral resource prediction platform: integration of multifunctional intelligent agents, mineral resources prediction platform, data and computational resource database, knowledge graph and large language models, and large predictive computing models