The ability to perform aeroacoustic measurements in closed-section wind tunnels is important in vehicle design, but presents significant problems compared with anechoic open-jet tunnel testing. The use of microphone arrays and beamforming processing can yield useful source localisation information, but there are concerns as to the consistency of quantitative results from such methods. In this paper some of the deficiencies of current beamforming practice are highlighted; in particular, inaccuracies in levels arising from the reverberant sound field in hard-walled wind tunnels. Notably, the use of free-space Green's functions when beamforming in such conditions is shown to be a significant source of error, and an image source method is proposed to improve accuracy by de-reverberating the measuring environment. Comparisons are made between the conventional and proposed approaches using microphone array measurements of a compact source in a closed section wind tunnel. Results show that the new de-reverberation method can give improvements in accuracy even if the wind tunnel geometry is not modelled accurately, provided that the number of image sources is chosen correctly. The results also highlight the significant challenges associated with Green's functions measurements in highly-reverberant environments.