This paper proposes a linear parameter varying (LPV) unknown input observer for the diagnosis of actuator faults and icing in unmanned aerial vehicles (UAVs). The accretion of ice layers on wings and control surfaces modifies the shape of the aircraft and alters the performance and controllability of the vehicles. The correct detection of this phenomenon is of paramount importance for the efficient implementation of de-icing techniques. The advantage of deriving the unknown input observer within the LPV framework is the possibility to deal with the nonlinearities of the UAV model by embedding them within some varying parameters. Results obtained with a Zagi Flying Wing simulator are used to validate the effectiveness of the proposed approach.