Accurate delineation of lake surface area is fundamental for understanding eco-hydrological processes in arid regions, yet long-term lake records are often constrained by cloud contamination, seasonal ice cover, and data gaps. In this study, we develop an optimized lake-area extraction framework that integrates seasonal water-index selection, adaptive threshold segmentation, maximum connected-component analysis, and mutual-information-based image gap filling to construct a continuous monthly lake-area time series for Bahannao Lake from 1984 to 2024. This framework substantially improves the temporal continuity and robustness of long-term monitoring for small lakes in arid environments, and its regional applicability is further validated through comparative analyses with Hongjiannao Lake and Wuliangsuhai Lake. Based on the reconstructed time series, we quantitatively assess the multi-climatic controls on lake-area variability by combining correlation analysis with an XGBoost model. The results reveal pronounced seasonal differences and distinct stage-dependent evolution in lake dynamics, with the dominance alternating between precipitation-driven water input and evaporative demand across different temporal scales. Our findings highlight the nonlinear hydro-climatic responses of arid-region lakes to climate variability and provide both technical support and scientific insight for long-term lake monitoring and water-resource management in dryland regions.