Spatiotemporal variations and overflow risk analysis of the Salt Lake in the Hoh Xil Region using machine learning methods

Researchers have identified that the Salt Lake in Qinghai Hoh Xil, following the 2011 Zonag Lake outburst, is rapidly expanding at a rate of approximately 350% across all directions, primarily toward the southeast. Without intervention such as an emergency drainage channel, machine learning models predict the lake's overflow could occur as early as 2020 and potentially as late as 2031. The study employed Classification and Regression Trees (CART), Random Forest (RF), and Support Vector Machine (SVM) algorithms to analyze spatiotemporal variations from 1973 to 2021, with RF demonstrating superior accuracy in lake area extraction (KAPPA = 0.98, overall accuracy = 0.99). The normalized difference water index was found to be a key feature variable. This research highlights the potential risks posed by such hydrological changes and underscores the effectiveness of machine learning in predicting environmental shifts.

Wang, R., Guo, L., Chen, B., Yang, Y., Zheng, H., Deng, F., & Liu, J. (2023). Spatiotemporal variations and overflow risk analysis of the Salt Lake in the Hoh Xil Region using machine learning methods. Frontiers in Earth Science, 10. https://doi.org/10.3389/feart.2022.1084540