As wind turbines are usually clustered in wind farms, knowledge of the evolution of the wind turbine wakes is important because downstream turbines will be exposed to them, causing higher loads and maintenance times. For that reason, wind turbine wakes have been studied intensively and different engineering wake models were derived. However, none of them is constructed from a few basic and robust assumptions, while such a formalism already exists for the axisymmetric turbulent wake. Therefore, we will apply these models to data obtained in a wind farm using a scanning LiDAR. The wakes of two wind turbines are analyzed in four different wind directions chosen so that their wakes will have different degrees of interaction. The axisymmetric wake models are found to perform better than the Jensen wake model, and the main source of improvement is shown to be the presence of a virtual origin. Indeed, a virtual origin is shown to also improve the Jensen wake model significantly. Overall, our results indicate that a view from classical axisymmetric wake theory can help to improve the understanding of the evolution of wind turbine wakes in the atmospheric boundary layer.