For the unmanned aerial vehicle (UAV) communication scenarios, the time-varying fading characteristics cased by the three-dimensional(3D) scattering environment and 3D trajectory have an important impact on establishing the reliable communication link. In this paper, a generic 3D non-stationary geometry-based stochastic model (GBSM) for UAV channels is developed. Different from the traditional 3D GBSMs, the proposed model takes the 3D arbitrary trajectories of UAV into account. On this basis, two important second-order statistical properties of fading envelope, i.e., the level crossing rate (LCR) and average fading duration (AFD), are investigated and derived in details. The obtained closed-form expressions are explicit functions of flight parameters and compatible with the results in the previous works. Numerical simulations show that the simulated results are in accordance with the theoretical and measurement data under 3D flight scenarios. The proposed model and statistical properties can be applied to the optimal design of channel coding and block interleave schemes for UAV communication systems. INDEX TERMS unmanned aerial vehicle (UAV) channel, channel model, fading envelope, level crossing rate (LCR), average fading duration (AFD). I. INTRODUCTION U NMANNED AERIAL VEHICLES (UAVs) have been attracting more and more attention in the military and civilian fields, such as weather monitoring, forest fire detection, filming, goods transport, and traffic control, for their affordable prices and high flexibility [1]. The UAV-assisted communications have been considered as an important option in the fifth generation (5G) communication systems [2], [3] and are also essential for extending wireless network coverage and relay communications [4], [5]. Due to the high-altitude and high-speed flight of UAV, the propagation environment of UAV communication is quite different from the conventional cellular and vehicular communications [6]. For example, the Doppler frequency changes rapidly due to the three-dimensional (3D) movement of UAV in the 3D space. This would lead to the severe fluctuation of received power and the deterioration of system performance [7], [8]. Therefore, a thorough understanding of UAV channel and its characteristics is vital to guarantee the high-speed transmission and reliability requirements of UAV communication systems. Particularly, it is a hot topic to investigate the statistical properties of UAV channels, such as probability distribution function (PDF), autocorrelation function (ACF), cross-correlation function (CCF), Doppler power spectral density (DPSD), level crossing rate (LCR), and average fade duration (AFD) [9]-[13]. Note that LCR and AFD are two important second-order statistical properties reflecting the fluctuation of fading envelope over time, and they have an important impact on the channel coding and block interleave schemes. However, the investigations of LCR and AFD for UAV channels are very limited compared with other statistical properties.