In a large eddy simulation (LES), resolving the wide spectrum of large turbulent eddies from [GRAPHICS] to [GRAPHICS] in the atmospheric boundary layer (ABL) requires [GRAPHICS] computational degrees of freedom; however, these eddies are intermittent in space and time. In this research, we take advantage of the spatial intermittency in a neutrally stratified atmospheric Ekman boundary layer, and study the development of a novel LES methodology. Using the second generation wavelet transform, the proposed model filters the large eddies into distinct groups of significant and insignificant eddies. We show that the significant eddies are sufficient to resolve the physics of the flow. The effects of insignificant eddies are modelled with the proposed multiscale parameterization scheme. The results of the proposed model have been found to be in good agreement with that of an equivalent reference model, experimental data, and asymptotic boundary layer theory. We have found that the number of significant eddies in a neutrally stratified ABL is much lower than the number of resolved eddies in a reference model. The overall algorithm is asymptotically optimal - the CPU time is approximately proportional to the number of resolved eddies. The proposed methodology suggests a potentially novel research direction that may be employed to address a number of computational challenges that must be faced in the field of atmospheric modeling.