Arid Central Asia (ACA), accounting for one-third of the world arid regions, is noted for its sparse freshwater resources and delicate ecosystems. Lakes in this region are sensitive to the complex interactions between climate warming, human impact, and the water cycle, affecting aquatic ecosystems as well as species distributions. Here, we reanalyzed relationships between measured environmental variables and the spatial distribution of 40 previously published lake surface sediment diatom (Bacillariophyceae) assemblages along with synthesized sediment cores from 4 distinct locations in Bosten Lake with the intention of developing a diatom-inference model for water depth. Diatom assemblage variations differed across surface sediment samples and within cores in Bosten Lake. The initial period (1770-1910 CE) is characterized by oligotrophic taxa such as Lindavia radiosa and Navicula peroblonga, which transitioned towards meso-eutrophic species such as Pseudostaurosira, Staurosira, and Staurosirella spp. (formerly classified in the genus Fragilaria) during 1910-1960 CE and notably lake became more eutrophic post-1960, marked by the abundance of Fragilaria cf. crotonensis. Redundancy analysis (RDA) showed that lake water depth explained a statistically significant portion of diatom assemblage variations. Weighted-averaging partial least squares (WA-PLS) is used to develop a diatom-based depth inference model (R2jack = 0.94, RMSEP = 1.08 m, and maximum bias = 2.57 m) based on diatom optima from Bosten Lake. Diatom-depth reconstructions and changes in diatom community composition are of great importance to track lake ecosystem dynamics, ensuring the sustainability of vital freshwater resources in the ACA and beyond.