In this paper, we propose a novel adaptive decoding mechanism (ADM) for the unmanned aerial vehicle (UAV)-enabled uplink (UL) non-orthogonal multiple access (NOMA) communications. Specifically, considering a harsh UAV environment where ground-to-ground links are regularly unavailable, the proposed ADM overcomes the challenging problem of conventional UL-NOMA systems whose performance is sensitive to the transmitter's statistical channel state information and the receiver's decoding order. To evaluate the performance of the ADM, we derive closed-form expressions for the system outage probability (OP) and throughput. In the performance analysis, we provide novel expressions for practical air-to-ground and ground-to-air channels while taking into account the practical implementation of imperfect successive interference cancellation (SIC) in UL-NOMA. These results have not been previously reported in the technical literature. Moreover, the obtained expression can be adopted to characterize the OP of various systems under a Mixture of Gamma (MG) distribution-based fading channels. Next, we propose a sub-optimal Gradient Descent-based algorithm to obtain the power allocation coefficients that result in maximum throughput with respect to each location on UAV's trajectory, which follows a random waypoint mobility model for UAVs. Numerical solutions show that the ADM significantly improves the performance of UAV-enabled UL-NOMA, particularly in mobile environments.