An advanced sigma-point Kalman filter was tested in a complex model and outperformed standard filters like EKF and EnKF. The SPKF method improved state and parameter estimation accuracy, especially in chaotic systems. It eliminated issues with tangent linear models and Jacobians, making it more effective for real-world ocean and atmospheric models. A reduced sigma-point subspace approach was proposed for higher-dimensional systems, showing consistent and better results compared to EKF and EnKF. The SPKF data assimilation scheme is a promising tool for accurately estimating states and parameters in strongly nonlinear dynamical models.