Flush air data sensing (FADS) systems have been successfully tested on the nose tip of large manned/unmanned air vehicles. In this paper we investigate the application of a FADS system on the wing leading edge of a micro (unmanned) air vehicle (MAV) flown at speed as low as Mach 0.07. The motivation behind this project is driven by the need to find alternative solutions to air data booms which are physically impractical for MAVs. Overall an 80% and 97% decrease in instrumentation weight and cost respectively were achieved. Air data modelling is implemented via a radial basis function (RBF) neural network (NN) trained with the extended minimum resource allocating network (EMRAN) algorithm. Wind tunnel data were used to train and test the NN, where estimation accuracies of 0.51°, 0.44 lb/ft2 and 0.62 m/s were achieved for angle of attack, static pressure and wind speed respectively. Sensor faults were investigated and it was found that the use of an autoassociative NN to reproduce input data improved the NN robustness to single and multiple sensor faults. Additionally a simple NN domain of validity test demonstrated how the careful selection of the NN training data set is crucial for accurate estimations.