High-Resolution Numerical Weather Simulation with a Large Domain for West Japan Extreme Heavy Rainfall Events during July 2018

A record-breaking rainfall occurred in west Japan in July 2018. The features of this rainfall event were that almost half of the main island of Japan was covered by rain area for few days and record break rainfalls were observed in many rain gauge stations. The rainfall caused severe floods and debris flow throughout western Japan. The damages of this rainfall were the largest since the Nagasaki flood in 1982. To mitigate these disasters damage and evacuate residents in a timely fashion, an accurate numerical weather prediction (NWP) system is important.

In this study, we simulated this heavy rainfall event using a high-resolution numerical weather prediction (NWP) model with a large model domain. The Japan Meteorological Agency Non-Hydrostatic Model and the K computer which was the flagship supercomputer in Japan were used for this huge simulation. Several important model factors (grid spacing, PBL scheme, terrain data and model domain size) influencing heavy rainfall forecasting in NWP models were investigated. One of the features of this study was that we simulated an operational model domain (3160*2600 km) with 500-m grid spacing.

The result showed the 2-km grid spacing models showed clearly better precipitation performance than 5-km resolution models. And the 500-m resolution models showed better precipitation performance than 2-km grid spacing models. The 2-km and 500-m grid spacing models with the large model domain show better performance than those with the small domain. Overall, the results indicated that using a high-resolution model (500-m grid spacing) with the large model domain area provides an advantage for simulating heavy rain events.