Initially, the WRF was run using 9-km, 3-km, and 1-km domains using the GFS-analysis for initial and boundary conditions for three OLYMPEX AR events. A few different microphysical schemes were used (P3, Morrison, and Thompson). The 1-km WRF runs underpredicted all events by 10-40%, and the underprediction was greater for Morrison and Thompson because of too little riming. It was hypothesized that the additional underprediction is the result of not resolving the turbulent eddies and smaller convective cells over the terrain that may contribute to additional accretional growth and fallout. Thus, the WRF was run nested down to 333-m and 111-m domains using the LES with the 1-km WRF as boundary conditions and the P3 microphysical scheme. During the blocked period, the low level jet (LLJ) was lifted by the terrain, with strong vertical wind shear layer above and below LLJ resulting in shear-driven turbulence. Meanwhile, convective cells were generated within a potentially unstable layer above the LLJ. The WRF-LES better resolved these convective cells and larger turbulent eddies. LES run shows the enhanced droplet nucleation, condensation and collision and coalescence processes in the turbulent region. The 111-m LES reduced the underprediction in the 1-km run by ~30% on the lowland region and ~40% on the windward sides. However, there still remains under-predicted precipitation, some of which may be from blocking front in WRF moving eastward too fast.