Abstract: To enhance the reliability of unmanned aerial vehicle (UAV) communication systems in low-altitude complex industrial environments and to support the development of low-altitude intelligent connectivity technologies and smart port construction, an intelligent channel model for UAV inspection scenarios at port bulk cargo terminals is proposed. Owing to complex shadowing from heavy machinery, random blockage from irregular cargo stacking, and altitude-dependent propagation conditions, UAV-to-ground (U2G) channels in this environment exhibit severe multipath effects, rapid time variation, and pronounced non-stationarity. Therefore, accurate characterization of channel properties in such complex port environments is required to ensure reliable communication links and stable cooperative operation of UAV swarms. A high-precision ray-tracing (RT) technique is adopted to construct a U2G communication dataset for a port bulk cargo terminal scenario, covering different yard utilization rates (40% and 70%), multiple frequency bands (5.9 GHz and 28 GHz), and multiple flight altitudes (80 m and 120 m). The constructed dataset captures key propagation mechanisms in port environments, including shadowing, scattering, and multipath effects induced by large metallic structures and irregular cargo piles. Based on the RT data, key statistical characteristics of the U2G channel are systematically analyzed, including the time autocorrelation function (TACF), singular value spread (SVS), and Doppler power spectral density (DPSD). The TACF is computed under different flight altitudes, communication frequency bands, and yard utilization conditions. The analysis indicates that, compared with the 5.9 GHz band, the TACF in the 28 GHz band exhibits a faster decay rate due to shorter wavelength and higher sensitivity to environmental blockage. TACF also decays faster at 120 m compared with 80 m, which is attributed to increased propagation distance and enhanced scattering effects. Furthermore, under high yard utilization conditions, dense cargo stacks generate a large number of scattering clusters, resulting in accelerated TACF decay and stronger channel time variation. In addition, SVS and its cumulative distribution function are derived through singular value decomposition of the channel matrices. The SVS distribution in the 28 GHz band is observed to shift toward larger values, reflecting increased channel sparsity and reduced spatial degrees of freedom. Higher UAV altitudes yield steeper SVS curves dominated by line-of-sight (LoS) component, whereas lower altitudes exhibit broader distributions caused by frequent shadowing. Moreover, high yard utilization significantly increases the SVS magnitude, indicating severe degradation of spatial multiplexing capability in dense port environments. DPSD analysis further reveals stronger Doppler frequency shifts at lower flight altitudes, higher frequencies, and higher yard utilization levels, confirming intensified channel non-stationarity. The results indicate that under severe blockage conditions at an 80 m altitude with a 70% yard utilization rate, non-line-of-sight (NLoS) paths increase sharply, and the multipath power proportion climbs to nearly 45%, exhibiting highly significant multipath fading effects. To accurately describe the observed channel behaviors, a geometric-based stochastic model (GBSM) is proposed for U2G communication channels in port bulk cargo terminal scenarios. This model incorporates a birth-death process and a cargo stacking factor to accurately characterize blockage, scattering, and multipath evolution under time-varying conditions imposed by large metallic machinery and irregular cargo piles. The capability of the proposed GBSM to characterize channel non-stationarity and statistical consistency is validated through comparison with RT data. The validation results demonstrate that the proposed GBSM achieves high accuracy in representing realistic port communication environments, thereby providing a reliable foundation for the design and optimization of UAV communication systems in smart ports and other complex industrial scenarios.