Exploring the Causes of Severe Fluctuations in Water Surface Area Using Water Index and Structural Equation Modeling: Evidence from Ebinur Lake, China

Arid zone lakes function as indicators of watershed ecology and environment, significantly influencing regional social development. In Ebinur Lake, a fuzzy water–land boundary hinders lake area extraction using remote sensing. Furthermore, unquantifiable anthropogenic–natural factors make it difficult to explore the drivers of lake area change. Utilizing Google Earth Engine (GEE), this study employs Landsat series, Sentinel 2, and MOD09GA/MYD09GA data to extract the water area of Ebinur Lake by applying indices such as NDWI, MNDWI, AWEI, and MAWEI. Threshold determination and shoreline refinement are achieved through Otsu’s method and the Canny algorithm, followed by a comparative analysis. Monthly spatiotemporal variations between 2009 and 2023 are analyzed using distance-level analysis and center-of-gravity analyses. It is noteworthy that this study adopted PLS-SEM. By comprehensively considering multifactorial interactions, this approach effectively simulates real-world natural scenarios and quantitatively evaluates the complex impacts of hydrology, meteorology, soil–vegetation, and human activities that influence changes in lake area. The results are as follows: (1) MAWEI outperforms NDWI, MNDWI, and AWEI with >95% overall accuracy and a Kappa coefficient >0.9, making it the best index for water body extraction; (2) from 2009 to 2017, Ebinur Lake’s area gradually increased, falling within a range of 450 km2 to 900 km2. Conversely, from 2017 to 2023, the lake’s area notably decreased, falling to between 330 km2 and 880 km2; (3) Ebinur Lake’s center of gravity shifts northwest to southeast, with primary changes in northwest mini-lake and transition zones; (4) hydrological factors were identified as the primary driver of changes in the Ebinur Lake area, contributing 64.3% of the total impact. Soil–vegetation, human activities, and meteorological factors contributed 16.7%, 11.3%, and 7.8%, respectively. The quantified driving factors and the MAWEI-based monitoring framework can directly provide references for water resource allocation policies and ecological restoration priorities in the economic zone of the Tianshan Mountains.