Abstract: To address the non-stationary characteristic of air-to-air (A2A) communications of unmanned aerial vehicles (UAVs) on sea surface, this paper proposes a channel modeling method based on a geometric-based statistical model (GBSM). In the sub-6 GHz frequency band, the channel model among UAVs comprehensively considers line-of-sight (LoS), sea surface specular reflection, and diffuse scattering paths, while integrating the JONSWAP wave spectrum to depict dynamic sea surface. On the basis of establishing channel impulse response (CIR), it deeply analyzes and quantifies parameters that describe channel time dispersion and frequency selective fading, including delay spread and coherent bandwidth. Meanwhile, it also analyzes the parameters that indicate the time-varying characteristics of channel, such as Doppler frequency shift and Doppler spread. Simulation results reveal the impact of different sea states on the time-frequency characteristic of signals and suggest the regulatory effect of marine environment on core parameters delay spread and Doppler spread of A2A link. The findings provide significant theoretical support for waveform design, parameter optimization, and anti-fading algorithm design in sea surface UAV integrated sensing and communication (ISAC) systems, contributing to the design and optimization of high-reliability communication in complex maritime environments.