[S-RD24] Cutting-edge sensing technology applied to geology and resource exploration

[SRD24-P03] Identification of Lithium Ore Distributed in the Gobi Desert of Mongolia Using Multispectral Drone

*Shion Wakae¹, Kaito Takizawa¹, Natsuo Okada¹, Tobimaru Ishiwata¹, Ryosuke Kikuchi¹, Yoko Ohtomo¹, Youhei Kawamura¹ (1.Hokkaido University)

Keywords:drone, lithium ore, Multispectral imaging, Hyperspectral imaging, Gobi desertert, Mongolia

Driven by the rapid expansion of renewable energy technologies-most notably electric vehicles (EVs) and energy storage systems-the global demand for lithium has experienced a significant surge. Consequently, new lithium deposits are being actively explored around the world. Among these, the Gobi Desert in Mongolia has emerged as a promising site for lithium exploration, owing to its lepidolite-bearing pegmatite deposits (K(Li,Al)₃(Si,Al)₄O₁₀(F,OH)₂), which are attracting attention for their low roasting costs. In this study, lepidolite ore exploration was conducted using drone photogrammetry equipped with a multispectral (MS) camera. Between September 18 and 21, 2024, a total of 440 aerial images were captured over a designated area where trenching surveys had exposed lithium ore. These images were processed to generate orthomosaic maps and MS data cubes, which were subsequently used to train a convolutional neural network (CNN) model for lithium ore detection. The developed system achieved an overall classification accuracy of 89.1%, with the lithium-rich regions mapped by the CNN showing strong agreement with field observations. Additionally, hyperspectral (HS) data were collected from sampled lepidolite to capture the full spectral characteristics of lithium ores. To improve the accuracy of lithium ore detection, we applied Neighborhood Component Analysis (NCA) to hyperspectral data collected from lepidolite samples. This analysis helped us identify ten key wavelengths that are most effective for distinguishing lithium-rich areas from surrounding rocks. Based on these findings, a refined six-band MS camera configuration is proposed-covering the RGB channels along with bands at 450 nm, 560 nm, 650 nm, 820 nm, and 920 nm-to further enhance the detection accuracy of lepidolite ore. This approach is expected to provide a sustainable and cost-effective strategy for lithium exploration in remote desert regions and to contribute significantly to future exploration efforts in Mongolia and beyond.