Based on observational data of annual maximum snow depth in Shihezi from 1960 to 2020, this study systematically investigates the temporal variation characteristics, abrupt change features, and multi-scale periodic patterns of snow depth, and conducts short-term forecasting. First, linear regression and mutation point detection methods are applied to analyze interannual trends and identify abrupt change years. Results show a significant increasing trend in snow depth, with an average growth of approximately 2.026 cm per decade. The year 2007 is identified as a notable mutation point, and the majority of snow depth values fall within the range of 10–35 cm. Second, Ensemble Empirical Mode Decomposition (EEMD) is employed to extract periodic components at different time scales, revealing significant cycles of approximately 10, 15, and 45 years. Finally, a short-term forecasting model is constructed to predict future trends. The projected annual maximum snow depths for 2026, 2027, and 2028 are 29.81 cm, 29.63 cm, and 32.19 cm, respectively. These findings provide scientific reference for regional snow hazard assessment, climate change impact analysis, and decision-making support.