Abstract: To understand the dynamic characteristics of underwater acoustic channels in transitional sea areas with undulating seabed topography and to comparatively analyze the estimation performance of different channel estimation algorithms in such environments, this paper investigates the characteristics of time-varying cluster-sparse underwater acoustic channels and adaptive time-varying underwater acoustic channel tracking methods. Using sea trial data for algorithm comparison, the performance and computational complexity of affine projection algorithms (APA), improved proportionate APA (IPAPA), and norm-constrained IPAPA in tracking rapidly time-varying channels are analyzed from perspectives such as Doppler shift, intra-cluster tap phase jitter and fluctuation, rapid tap amplitude variation, and large-scale attenuation. The relationship between phase/amplitude variations of cluster taps and fluctuations in channel estimation performance is established. Analysis reveals that nonlinear Doppler drift of taps and large-scale amplitude attenuation lead to degradation in channel estimation performance. It is demonstrated that the IPAPA algorithm achieves a balanced performance in terms of symbol mean square error and hardware feasibility, laying the foundation for channel evaluation in complex marine environments and the prediction of underwater acoustic communication quality.