Abstract: Single-model bridge backwater calculation methods have reputations of poor error controllability and low computational efficiency when applied to complex terrain. To address these issues, this paper proposes a multi-model fusion computational framework based on a task division-feedback-integration architecture. The framework integrates the HEC-RAS one-dimensional model, the MIKE 21 two-dimensional hydrodynamic model, and three empirical formulas for backwater calculation. A dynamic data assimilation mechanism enables parameter interaction and error suppression among the models. The Bayesian model averaging (BMA) method is introduced to adaptively optimize weights, thereby improving backwater prediction accuracy while maintaining high computational efficiency. Taking six bridges in the Zhuoni section of the Taohe River as a case study, the proposed framework was verified under different flood recurrence frequency conditions. The results show that, through comparative analysis of three testing methods, the fusion framework exhibits the smallest error fluctuation range and the highest comprehensive accuracy across various flood frequencies and pier types. The computation time of the fusion framework is only 10.5 minutes, which is more than 30% faster than that of a single numerical simulation. The fusion framework avoids the blindness of empirical formulas and the limitations of single numerical models, providing more reasonable estimates and clearer mechanisms for backwater length prediction. This not only prevents engineering waste caused by overestimation but also mitigates flood control safety hazards due to underestimation. When key river channel parameters are disturbed, the fusion framework achieves the lowest maximum relative error and demonstrates good adaptability to heterogeneous pier structures. Overall, the proposed fusion framework offers a high-precision and high-efficiency calculation method for bridge flood control safety assessment, providing technical support for bridge backwater analysis and engineering optimization in complex river environments.