Salt crusts formed on inland arid playas are an important global source of saline dust, posing severe risks to ecosystem stability and human health. Understanding salt crust dynamics and driving factors is essential for assessing saline dust emissions and ecological risks. The Ebinur Lake Basin (ELB) is a significant source of saline dust in northwest China and features extensive salt crusts. However, the spatiotemporal evolution of salt crusts has not been well quantified due to the lack of consistent long-term observations. In this study, an improved nonlinear spectral unmixing model based on ensemble tree models is developed to estimate salt crust coverage in the ELB spanning 2005-2025. The model performs robustly (R2 = 0.769) for medium-to-high coverage salt crusts (MHSCs; fractional cover >= 0.3). From 2005 to 2025, the MHSC area in the ELB increased with interannual fluctuations, peaking at 377.8 km2 in 2021-approximately 5.03 times the 2005-2009 mean. Spatially, the MHSCs were categorized into four types: lake-basin, desert, fluvial-depressional and anthropogenic. Among these, the lake-basin type covered the largest area and exhibited the strongest variability, accounting for most of the interannual fluctuations in MHSC area. Analysis of the driving factors indicates that lake area, drought conditions, and human activities were the primary drivers of the MHSC changes, with drought having a distinct lagged effect. This study estimates the long-term spatiotemporal dynamics of the MHSCs in the ELB and provides a scientific basis for the ecological risk assessment and the management of saline dust hazards in arid and semiarid regions.