Verification of the performance of the high resolution QPF model for heavy rainfall event over the Korean Peninsula

The coupled atmospheric (global and local)-diagnostic rainfall model (called QPM) simulation is performed with various resolutions in order to discover the most appropriate nesting process in simulating a heavy rainfall event which led to an extreme amount of rainfall and flash flood in July 2008 over the Korean Peninsula. A series of experiments consists of six QPM simulations given 40 and 20 km global runs, two 7.8 km local runs from single nesting and other two 2.2 km local runs from double nesting. Four verification approaches focused on accuracy and efficiency are carried out in order to evaluate the coupled model system performance. The results show that the QPM simulated total accumulated rainfall from 20 km global run or 7.8 km local runs successfully captures the observed rainfall over a spatial observation network. Furthermore, the evaluation of the peak rainfall amount and the rainfall area over a given time interval demonstrates that the QPM forecast from either 20 km global run or 7.8 km local runs shows the best agreement with observation. In addition, the quantitative evaluation of the model performance by computing simple statistical measures presents the good agreement between the simulations and the observation, especially when the QPM forecasts are produced by the 7.8 km resolution local model. Finally, the modeling system which couples 40 km global-7.8 km local models and the diagnostic rainfall model is proved to be the best nesting approach in the natural disaster prediction system when considering the accuracy and efficiency. However, the verification for the nesting processes over the long-term period still deserves to be studied for a successful prediction system.