Identification of spodumene using a remote-sensing index cube from SDGSAT-1 and other satellites

Abstract

Lithium (Li), a key element in green energy technologies, plays a pivotal role in achieving the United Nations Sustainable Development Goals (SDGs). Globally, spodumene-bearing pegmatite Li deposits are the primary source of Li. In this study, a series of spodumene thermal infrared indices (STIRI) were derived via linear regression using thermal infrared (TIR) data from the SDGSAT-1, ASTER, and Landsat 8 satellites. These STIRI were combined with a band ratio (BR) index from the Landsat 8 visible near-infrared (VNIR) bands, and the matched filter (MF) index from the GF-5 shortwave infrared (SWIR) bands to construct a remote-sensing index cube. This cube was fed into a hybrid deep-learning model to identify spodumene in the 509 Li deposit in Dahongliutan, Xinjiang, China. The model combines a convolute onal neural network (CNN) and a graph convolutional network (GCN), integrating spatial and spectral features to enhance identification accuracy. The results showed a significant improvement in identification accuracy when TIR indices were incorporated alongside the VNIR and SWIR indices. Notably, SDGSAT-1 TIR data, with a 30 m spatial resolution and broader spectral range, proved highly effective in spodumene detection, offering new opportunities for Li ore prospecting in high-altitude, deeply eroded regions.

Keywords

• Remote sensing
• SDGSAT-1
• thermal infrared
• spodumene identification
• deep learning