Traditionally buildings are designed assuming they will be loaded by a statistically stationary atmospheric boundary layer wind with a logarithmic mean vertical profile. However, there are other wind types which differ from this. For example, the thunderstorm downburst is highly non-stationary and has a different vertical velocity profile. This presents a problem as existing wind tunnels, analysis techniques, scalings and assumptions about wind loading may be incorrect for the downburst. To address these issues a pulsed impinging jet simulator was developed. The flow field was scaled and then compared to a full scale event using non-stationary analysis parameters based upon wavelet analysis. When scaled to a medium intensity downburst the simulator had scales of: length, 1 : 1000, velocity 1 : 1.67 and time 1 : 1109. However, the scalings were not self-consistent, suggesting it was only capable of a partial simulation, i.e. it could only simulate part of the velocity time history of a full scale down burst. Pressure, force and lift coefficients were then calculated for a model CAARC building placed in the simulator at the location of maximum velocity. A comparison to ABL results for a single and interference effects case revealed that, in the single building case, the downburst wind loads exceeded the ABL case near the base of the building, where wind speeds exceeded that of the ABL. The interference case was more complex, generally interference effects were reduced in downburst flows. However, at certain yaw angles and separation distances exceedances over the single building case occurred. Further research is still needed into the interference effects phenomenon in downburst flows.