Air-to-ground (A2G) communications links are an important research area for non-terrestrial networks, which are mostly included in the standards together with 5th generation (5G) communications. In this study, a novel approach has been developed to dynamically determine whether multi-carrier or single-carrier waveforms will be a better choice under varying environmental conditions for A2G communications links, using machine learning techniques. Within the framework of this developed approach, environmental information affecting the A2G communications channel are used as the features. Different machine learning models are compared with each other on an artificially generated dataset.