Prediction of confined water level within the levee through long-term water level record-based probabilistic modeling

This study aims to develop a prediction method based on long-term water level observation data to provide a basis for determining the anti-floating safety water level within T. Utilizing long-term groundwater level observation records and hydrological data outside the levee from a project case within the Beijiang levee, the confined water level was taken as the study object. Probability modeling of observation records was implemented through the empirical distribution method and extreme value distribution method, extrapolating the probability density function of the maximum confined water level within T to achieve prediction of long-term peak confined water levels. Using the 50-year return period confined water level standard value as the predictive indicator, the probability model yielded standard values of 10.32 m (empirical distribution) and 10.26 m (extreme value distribution) when calibrated against the 100-year return period confined water level derived from inversion/numerical methods. Conversely, when the probability model was established exclusively from observation records, the extreme value distribution approach gave a standard value of 9.96 m. Based on the standard values and observational analysis results, this study developed determination methods for the anti-floating safety water levels during both service and construction periods, while providing load statistical parameters for calibrating partial factors and combination factors for loads. By adopting probabilistic load modeling as the entry point, this research resolves critical issues in anti-floating design and establishes a basis for applying and optimizing reliability analysis. The technical approach demonstrates strong compatibility and implementability with significant potential for broader application.