This paper presents a method for characterising the primary dynamics of a rotary unmanned aerial vehicle. Based on first principles and basic aerodynamics, a mathematical model which explains the rigid body dynamics of a model-scale helicopter is developed. This model is reduced to three simplified decoupled models of attitude dynamics. Empirical test data is collected from a field experiment with significant wind disturbances. The method worked accurately on both uncoupled and fully coupled attitude models. An integral based parameter identification method is presented to identify the unknown intrinsic helicopter parameters as well as model of wind disturbance. An extended Kalman filter system identification method and common nonlinear regression are used for comparison. The EKF was found to be highly dependent on the initial states, so is not suitable for this application which contains significant disturbance and modelling errors. Nonlinear regression proved to be sufficiently accurate but computationally expensive. The proposed integral based parameter identification method was shown to be fast and accurate and is well suited to this application.