Abstract: Underwater target tracking is a crucial component of marine exploration missions. Due to complex underwater environments and target motion model switching, unknown statistical characteristics of noise arise, while employing fixed observation noise significantly degrades filtering accuracy. To address the unknown statistical properties of underwater observation noise and model switching delays, an adaptive interactive multiple-model square-root cubature Kalman filter incorporating the Sage-Husa noise estimation method is proposed and applied to a stationary dual-observation system. Initially, the square-root cubature Kalman filter (SRCKF) is utilized to estimate the target state, while incorporating a time-varying Sage-Husa noise estimator to adaptively estimate observation noise and correct the state estimates in real time. Meanwhile, within the interactive multiple model (IMM) algorithm, the transition probability matrix was updated by utilizing the rate of change in model probabilities at adjacent time steps, thereby improving the match between motion models and the actual trajectory. Simulation results show that the proposed adaptive IMM-SRCKF improves model-switching speed and adaptively estimates the statistical characteristics of the observation noise. In variable- noise environments, the proposed algorithm achieves superior tracking performance and outperforms conventional nonlinear filtering methods.