Unravelling mystery of lake color changes on the Tibetan Plateau: From mechanisms to prediction

• Machine learning models are used for nonlinear relationship related to lake color. • Lakes color changes on the Tibetan Plateau were studied by 12 environmental factors. • Lake geographic patterns show high importance with nonlinear relationship thresholds. • Future lake color is predicted under SSP1, SSP2, and SSP5. Lake color plays a crucial role in water quality monitoring, watershed ecological environment, tourism, and aesthetics, particularly in plateau areas. However, the patterns and mechanisms of lake color changes are still poorly understood. This study analyzed the seasonal and spatial changes of lake color (represented by Forel-Ule water color index changes) and employed machine learning models to evaluate the nonlinear relationship between environmental factors and lake color changes on the Tibetan Plateau. Then predictions were made for changes under different Shared Socioeconomic Pathways (SSPs). The results revealed that the lowest Forel-Ule was observed in spring of 6.65, followed by summer of 6.57 and autumn of 6.47. The top three most important factors were average lake water depth, average lake size, and average lake elevation. When these factors were respectively exceed 9.6 m, 110.89 km 2 , and 4675.0 m, the Forel-Ule tended to decrease. Additionally, lake water storage, water salinity, O 3 , precipitation, and temperature, as well as CO 2 , PM 2.5, and vapor pressure deficit, were also vital in determining Forel-Ule. Overall future decreasing trend was found under SSP1, SSP2, and SSP5. Lakes on the Yellow, Southern Tibet, and Western Tibet basins exhibited low Forel-Ule, with the change rate of −11.2%. Our findings provide the foundation for explaining lake color changes on the Tibetan Plateau and offers valuable insights for enhancing the aesthetic appeal of the landscape and aquatic life.