In this paper, we examine the Bayesian Cramér-Rao lower bounds (BCRBs) for channel estimation in an amplify-and-forward (AF) one-way relay network (OWRN) under time selective flat fading scenario, where the superimposed training is adopted at relay node in order to achieve individual channel estimation. We formulate the nonlinear dynamic state space for individual channels and derive online/offline BCRBs for partially-data-aided (PDA) estimator, which has imperfect statistical information about the data of the source, and possesses fully information of the symbols superimposed by the relay. we provide one framework to numerically calculate online/offline BCRBs. Finally, numerical results are provided to corroborate the proposed studies.