Hyperspectral remote sensing of white mica: A review of imaging and point-based spectrometer studies for mineral resources, with spectrometer design considerations
Hyperspectral remote sensing of white mica: A review of imaging and point-based spectrometer studies for mineral resources, with spectrometer design considerations
复制标题
DOI：
10.1016/j.rse.2022.113000
复制
发表时间：
2022-06
影响因子：
13.5
通讯作者：
John M. Meyer;R. Kokaly;E. Holley
中科院分区：
文献类型：
--
作者：
John M. Meyer;R. Kokaly;E. Holley
近30年来，白色云母化学变化的高光谱遥感已被证明可用于一系列存款类型的存款研究。为了更好地了解矿床和指导分光计设计，本文综述了遥感、光谱学和地质学领域利用白色云母化学变化引起的光谱变化的相关论文。这篇文章回顾了在以下类型的矿床中进行的光谱研究：贱金属硫化物、浅成热液、斑岩、沉积岩型金矿床、造山型金、氧化铁铜金和与不整合有关的铀。本文给出了白色云母的结构、化学组成和光谱特征。经审查的实验室光谱研究确定，白色云母2200 nm组合特征的位置偏移为1 nm，对应于Alot含量的变化约为± 1.05%。许多回顾的光谱研究表明，白色云母2200 nm组合特征的位置移动1 nm具有地质意义。带通、采样间隔和通道位置，使用具有深吸收特征的19个白色云母的光谱来进行，以确定精确测量云母位置的偏移所需的最小特征。白色云母2200 nm组合特征。确定需要< 16.3 nm的采样间隔和<17.5 nm的带通来实现2 nm的均方根误差（RMSE），而需要< 8.8 nm的采样间隔和<9.8 nm的带通来实现1 nm的RMSE。为了比较，常用的成像光谱仪HyMap、AVIRIS-Classic、SpecTIR®的AisaFENIX 1 K和HySpextmSWIR 384在确定2200 nm白色云母组合特征的位置时具有2.1、1.2、0.96和0.95 nm的RMSE，进行额外的灵敏度分析以确定信噪比（SNR）使用具有深吸收特征的18个白色云母的光谱，对白色云母2200 nm组合特征的位置进行RMSE。对于一个光谱仪的采样间隔和1 nm的带通，我们估计，RMSE为1和1.5 nm的光谱具有约246和64的最小信噪比，分别是可实现的。对于一个光谱仪的采样间隔和5 nm的带通，我们估计，RMSE为1和1.5 nm的光谱具有约431和84的最小信噪比，分别可达到。当使用采样间隔为8.8 nm且带通为9.8 nm的光谱仪时，RMSE为1仅可通过卷积的无噪声参考光谱实现。对于8.8_9.8 nm光谱仪，SNR为250和100的光谱分别导致RMSE为1.1和1.3。因此，精细的光谱分辨率特性对于高SNR光谱实现了优于1 nm的RMSE，而具有粗糙光谱分辨率的光谱仪具有较大的RMSE，对于噪声数据表现良好，并且如果1.1至1.5 nm的RMSE是可接受的，则对于白色云母研究是有用的。
Over the past ~30 years, hyperspectral remote sensing of chemical variations in white mica have proven to be useful for ore deposit studies in a range of deposit types. To better understand mineral deposits and to guide spectrometer design, this contrib ution reviews relevant papers from the fields of remote sensing, spectroscopy, and geology that have utilized spectral changes caused by chemical variation in white micas. This contribution reviews spectral studies conducted at the following types of mineral deposits: base metal sulfide, epithermal, porphyry, sedimentary rock hosted gold deposits, orogenic gold, iron oxide copper gold, and unconformity-related uranium. The structure, chemical composition, and spectral features of white micas, in this contribution defined as muscovite, paragonite, celadonite, phengite, illite, and sericite, are given. Reviewed laboratory spectral studies determined that shifts in the position of the white mica 2200 nm combination feature of 1 nm correspond to a change in Aloctcontent of approximately ±1.05%. Many of the reviewed spectral studies indicated that a shift in the position of the white mica 2200 nm combination feature of 1 nm was geologically significant.A sensitivity analysis of spectrometer characteristics; bandpass, sampling interval, and channel position, is conducted using spectra of 19 white micas with deep absorption features to determine minimum characteristics required to accurately measure a shift in the position of the white mica 2200 nm combination feature. It was determined that a sampling interval < 16.3 nm and bandpass <17.5 nm are needed to achieve a root mean square error (RMSE) of 2 nm, whereas a sampling interval < 8.8 nm and bandpass <9.8 nm are needed to achieve a RMSE of 1 nm. For comparison, commonly used imaging spectrometers HyMap, AVIRIS-Classic, SpecTIR®'s AisaFENIX 1K, and HySpextmSWIR 384 have 2.1, 1.2, 0.96, and 0.95 nm RMSE in determining the position of the 2200 nm white mica combination feature, respectively.An additional sensitivity analysis is conducted to determine the effect of signal to noise ratio (SNR) on the RMSE of the position of the white mica 2200 nm combination feature, using spectra of 18 white micas with deep absorption features. For a spectrometer with sampling interval and bandpass of 1 nm, we estimate that RMSEs of 1 and 1.5 nm are achievable with spectra having a minimum SNR of approximately 246 and 64, respectively. For a spectrometer with sampling interval and bandpass of 5 nm, we estimate that RMSEs of 1 and 1.5 nm are attainable with spectra having a minimum SNR of approximately 431 and 84, respectively. When using a spectrometer with a sampling interval 8.8 nm and a bandpass of 9.8 nm, a RMSE of 1 is only achievable with convolved, noiseless reference spectra. For the 8.8_9.8 nm spectrometer, spectra with SNR of 250 and 100 result in RMSE of 1.1 and 1.3, respectively. Therefore, fine spectral resolution characteristics achieve RMSEs better than 1 nm for high SNR spectra while spectrometers with coarse spectral resolution have larger RMSE, perform well with noisy data, and are useful for white mica studies if RMSE of 1.1 to 1.5 nm is acceptable.