We propose a method to detect icing of the airfoil of a fixed-wing Unmanned Aerial Vehicle by using an aerodynamic coefficient estimator and ambient temperature and humidity sensors. The estimator uses the information provided by a standard autopilot sensor suite consisting of an IMU, GNSS and a pitot-static tube to estimate lift coefficients as well as steady and turbulent wind velocities. These sensor inputs are fused within an Extended Kalman Filter using frequency separation and kinematic, aerodynamic and wind models while avoiding the need for prior knowledge about the aircraft. Ambient temperature and humidity sensors are used to assess environmental conditions and if icing is suspected, a trigger signal to the estimator and the autopilot is generated. This signal is used to adjust the anticipated uncertainties of the estimated coefficients and, if permitted by the flight control system, to initialize a small altitude change in order to excite the estimator. Simulation results show that this method is able to separate clearly between iced and non-iced cases and can be used to significantly enhance the detection performance compared to only using temperature and humidity based information.