Evaluation of Heavy Rainfall Model Forecasts over the Korean Peninsula Using Different Physical Parameterization Schemes and Horizontal Resolution

In this study, the accuracy of a Pennsylvania State University–National Center for Atmospheric Research mesoscale model （PSU/NCAR MM5） for predicting heavy summer precipitation over the Korean Peninsula was investigated. A total of 1800 simulations were performed using this model for 30 heavy rainfall events employing four cumulus parameterization schemes （CPS）, two grid-scale resolvable precipitation schemes （GRS）, and two planetary boundary layer （PBL） schemes in three model resolutions （90 km, 30 km, and 10 km）. The heavy rainfall events were mesoscale convective systems developed under the influence of mid-latitude baroclinic systems with low-level moisture transport from the ocean. The predictive accuracy for maximum rainfall was approximately 80% for 10-km resolution and was 60% for 30-km resolution. The predictive accuracy for rainfall position extended to ～150 km from the observed position for both resolutions. Simulated rainfall was most sensitive to CPS, then to PBL schemes, and then to GRS. In general, the Grell （GR） scheme and the Anthes and Kuo （AK） scheme showed a better prediction capability for heavy rainfall than did the Betts-Miller （BM） scheme and the Kain-Fritsch （KF） scheme. The GR scheme also performed well in the 24-h and 12-h precipitation predictions： the parameterized convective rainfall in GR is directly related to synoptic-scale forcing. The models without CPS performed better for rainfall amounts but worse for rainfall position than those with CPS. The MM5 model demonstrated substantial predictive capacity using synoptic-scale initial conditions and lateral boundary data because heavy summer rainfall in Korea occurs in a strong synoptic-scale environment.