Considering the uncertainty in communication and sampling in underwater target positioning process, it is crucial to calculate the variance lower bound of underwater target positioning error for evaluating performance, as the additional errors introduced by the solution of noisy networks are superimposed. In response to the problem of unstable positioning accuracy in underwater target maneuvering operation, Unscented Particle Filtering (UPF) and extended posterior Cramer-Rao Lower Bound (PCRLB) are adopted to carry out modeling and evaluation of approximate optimal accuracy for underwater target long endurance. Firstly, considering that multi parameter and nonlinear models are prone to filter degradation, the UPF is used to estimate the underwater target state. Secondly, Taylor series expansion is used to Solve approximate approximation of predicted states of underwater targets. Then, the state values by filtering estimation the expected and variance by approximate estimation are integrated into three-dimensional extended PCRLB. Finally, the performance comparison is conducted among the Unscented Kalman Filtering (UKF) from order 0 to order 2, Particle Filter (PF), Minimax Particle Filter (MPF), UPF from order 0 to order 2, and the theoretical optimal estimation. The results indicate that the approximate PCRLB estimated values of the target position and velocity for the long-term of the proposed model can approach the theoretical value, and thus can be used for the analysis of the long-term positioning performance of underwater targets.