秦祁昆造山带花岗伟晶岩型锂矿岩矿光谱特征及找矿应用  

Spectral Characteristics and Prospecting Applications of Granite Pegmatite Type Lithium Deposits in the Qinling,Qilian and Kunlun Orogenic Belts

作　　者：周斌[1,2] 高景刚[1,3] 梁婷[1,3] 余福承 李积清[4] 李永[4] 凤永刚[1,3] 张莉 Zhou Bin;Gao Jinggang;Liang Ting;Yu Fucheng;Li Jiqing;Li Yong;Feng Yonggang;Zhang Li(College of Land Engineering,Chang'an University,Xian,Shaanxi,710054,China;College of Earth Sciences and Land Resources,Chang'an University,Xian,Shaanxi,710054,China;Laboratory of Mineralization and Dynamics,Chang'an University,Xian,Shaanxi,710054,China;Qinghai Geological Survey Institute,Xining,Qinghai,810012,China)

机构地区：[1]长安大学地球科学与资源学院,陕西西安710054 [2]长安大学土地工程学院,陕西西安710054 [3]长安大学成矿作用及其动力学实验室,陕西西安710054 [4]青海省地质调查院,青海西宁810012

出　　处：《新疆地质》2025年第1期174-181,共8页Xinjiang Geology

基　　金：国家重点研发计划战略性矿产资源开发利用专项我国西部伟晶岩型锂等稀有金属成矿规律与勘查技术项目(2021YFC2901902)资助。

摘　　要：本文在柴达木盆地北缘东段、西昆仑大红柳滩及阿尔金3个地区系统采集了花岗伟晶岩样品,利用ASD光谱仪分别测量伟晶岩单矿物、岩石样品、岩石粉末样及围岩的光谱信息,分析了粉末样品Li含量。引入竞争性自适应加权重采样法(CARS)共选出107个锂含量敏感波段,基于敏感波段建立三维波段光谱指数(TBI),其中在1248 nm、1387 nm及1402 nm 3波段建立的TBI5指数与锂含量相关性系数达到0.66。结合偏最小二乘法(PLS)、随机森林(RF)和极限学习机(ELM)3种方法建立伟晶岩锂含量反演模型,基于107个敏感波段和TBI指数的RF法锂含量反演方法,实测光谱验证精度最高(R2=0.89),为利用伟晶岩的光谱信息寻找锂矿提供了新的思路。This study systematically collected granite pegmatite samples from the eastern part of the northern margin of the Qaidam Basin,the Dahongliutan area of the West Kunlun Mountains,and the Altyn Tagh.The ASD spectrometer was used to measure the spectral information of single minerals,rock samples,rock powder samples,and the surrounding rocks of the pegmatite,and the lithium(Li)content of the powder samples was analyzed.By introducing the Competitive Adaptive Reweighted Sampling(CARS)method,107 lithium-sensitive bands were selected.Based on these sensitive bands,a three-dimensional band spectral index(TBI)was developed,with the TBI5 index—constructed from the 1248,1387,and 1402 nm bands—showing a correlation coefficient of 0.66 with lithium content.A lithium content inversion model for pegmatite was established using partial least squares(PLS),random forest(RF),and extreme learning machine(ELM)methods.Among these,the RF-based lithium content inversion model,which used 107 sensitive bands and the TBI index,demonstrated the highest accuracy in spectral validation(R2=0.89).This provides a new approach for using the spectral information of pegmatite to explore lithium deposits.

关 键 词：花岗伟晶岩 秦祁昆造山带 光谱特征 锂 敏感波段 

分 类 号：P588.13[天文地球—岩石学] P618.71[天文地球—地质学]