Closed-section hard-walled wind tunnels are routinely used for aerodynamic testing during the early stages of vehicle design. The use of microphone arrays and beamforming processing in such environments can yield useful source localisation information; yet there are concerns as to the consistency of quantitative results from such methods when compared to anechoic open-jet tunnel testing. Higher background noise levels, microphones in contact with the turbulent boundary layer, and the highly reverberant nature of the measuring environment are of particular concern. In this paper we show that accurate results in closed-section wind tunnels are still possible using the cross-spectral matrix diagonal removal (DR) technique, and with de-reverberation. With DR, improvements in beamformer's accuracy up to 10 dB can be achieved when compared to standard beamforming. De-reverberation gives an additional 10 dB improvement, and provided that the noise contamination to the microphone pressure data is somewhat suppressed, errors within ±1dB can be expected. Both of these methods do not require any additional changes to the physical infrastructure of the measuring environment.