Unmanned aerial vehicles (UAVs) have gained popularity in the communications research community because of their versatility in placement and potential to extend the functions of communication networks. However, there remains still a gap in existing works regarding detailed and measurement-verified air-to-ground (A2G) Massive Multi-Input Multi-Output (MaMIMO) channel characteristics which play an important role in realistic deployment. In this paper, we first design a UAV MaMIMO communication platform for channel acquisition. We then use the testbed to measure uplink Channel State Information (CSI) between a rotary-wing drone and a 64-element MaMIMO base station (BS). For characterization, we focus on multidimensional channel stationarity which is a fundamental metric in communication systems. Afterward, we present measurement results and analyze the channel statistics based on power delay profiles (PDPs) considering space, time, and frequency domains. We propose the stationary angle (SA) as a supplementary metric of stationary distance (SD) in the time domain. We analyze the coherence bandwidth and RMS delay spread for frequency stationarity. Finally, spatial correlations between elements are analyzed to indicate the spatial stationarity of the array. The space-time-frequency channel stationary characterization will benefit the physical layer design of MaMIMO-UAV communications.