The Zabuye Salt Lake in Tibet, China is the only salt lake in the world that contains natural crystalline lithium carbonate. The grade and spatial distribution of mineral resources are of great importance to the development and utilization of salt lake mineral resources. The use of remote sensing technology for salt lakes observations can overcome the disadvantages of traditional station observations, such as spatial discontinuity, high time consumption, and high labor costs. In addition, machine learning algorithms can efficiently analyze the information from remote sensing data. In this study, Landsat-8 remote sensing image data and the Light Gradient Boosting Machine (LightGBM) algorithm were used to perform inversions of the depth, salinity, and lithium concentration of the Zabuye Salt Lake. Moreover, the water volume, total salinity, and total lithium content of Zabuye Salt Lake in 2000 and 2017 were estimated, and the distribution of mineral resources and changes during the study period were analyzed. The results show that the water depth and volume of the entire lake increased sharply in 2017, resulting in a decrease in salinity and lithium concentration in the lake. Due to the inflow from the surrounding dry salterns, the South Lake experienced a relatively small change. Furthermore, the amount of lithium resources in North Lake decreased significantly in 2017 compared to 2000, possibly due to higher temperatures during the month of observation, which led to precipitation of lithium carbonate. Our study proves the feasibility and accuracy of the LightGBM machine learning algorithm for rapid inversion of salt lakes, which provides technical insight into remote sensing inversion of other mineral resources in salt lakes. Thus, the development of remote sensing technology in recent years can provide increasingly detailed assessments of salt lake resources in the future.
Read full abstract