Local Analysis and Prediction System (LAPS) by FSL is applied to produce initial field of mesoscale numerical model for simulation of heavy rainfall event occurring over South Korea in July 1998. The reanalysis data from JMA is used as first guess, and GTS data, surface observation data in South Korea, and GMS brightness temperature are used in LAPS process. Numerical mesoscale model used in this study is well-known MM5 by PSU/NCAR. The heavy rainfall case chosen here has marked highest hourly precipitation (145mm/hr) ever in South Korea. Experiment with conventional pre-process (control run) and one with LAPS process are conducted to assess whether the LAPS has an considerable effect on increasing rainfall amount. We found a good improvement of 50% increase in rainfall amount in experiment with LAPS, which is related to enhanced mesoscale low system and low level wind field around the rainfall area.