Compatibility of High Resolution Terrain with High Resolution Model Grid

As computing power increases, mesoscale models are being run at a higher and higher resolution. Naturally, the user wants to represent the underlying terrain in the mesoscale model as realistically as possible by matching the terrain resolution to that of the model. However, from a theoretical standpoint, this may not be the best option. Skamarock (2004) showed that the effective resolution of ARW-WRF (Advanced Research Weather Research and Forecast) model is 7 Δx. Terrain features of 7 Δx or smaller may impose forcings on the WRF model that are unphysical or noisy. Typically, artificial smoothing is applied to terrain data to alleviate this problem. A question comes up with this procedure: what kind of and how much smoothing should one apply since smoothing decreases the realism of the terrain and defeats the purpose of running a high resolution model. In this paper, the flow over complex terrain in the Korean Peninsula will be examined using WRF to study this issue using discrete wavelet transform (DWT). WRF is set with two identical domains of the same horizontal extent with horizontal grid spacing of 3 and 9 km. Sensitivity experiments were conducted with the underlying terrain smoothed using DWT by removing the 2 Δx, 4 Δx, and 8 Δx terrain features. Sensitivity experiments were compared with each other to show how the model grid spacing and terrain smoothing affected phenomena such as sea/land breeze and flow around/over mountains. For the purpose of contrast and comparison, two case studies were chosen with different weather regimes. The first case occurred from 28-29 May 2012 with calm weather conditions in the Korean Peninsula. The second case occurred from 7-8 December 2012 when a cold front pushed southward into the Korean Peninsula. The ultimate goal of this study is to find a balance between smoothing the underlying terrain while maintaining realism of flow over complex terrain.