[A-AS08] General Meteorology

[AAS08-P05] Development of data assimilation techniques for GNSS wet delay from slant-path observation in complex terrain environments and verification of rainfall prediction.

*Shimizu Shingo¹, Kei Akama², Yuki Sato³ (1.National Research Institute for Earth Science and Disaster Resilience, 2.Information Technology Research and Development center, Mitsubishi Electric, 3.Advanced Technology Research and Development center, Mitsubishi Electric)

Keywords:Data assimilation of slant-wise wet delay, Data selection in complex terrain environments

The assimilation technology of water vapor using satellites is crucial for improving the accuracy of forecasts for heavy rain events, such as senjo-kousuitai, but developing it to be applicable even in complex terrain environments is an urgent issue. When calculating the slant delay amount, in complex terrain conditions such as those in the Kyushu region, it is necessary to 1) select available satellite data after considering terrain-induced obstructions in the observation PATH between ground stations and satellites, and 2) develop data assimilation that takes into account the difference between the elevation in numerical models and the actual elevation. In this study, focusing on the senjo-kousuitai that occurred in the Kyushu region on August 12, 2021, we calculated the slant wet delay amount addressing the aforementioned two points and conducted a data assimilation experiment. Cloud-resolving numerical model CReSS was used, conducting experiments at a horizontal resolution of 1km. For data assimilation, CReSS-3DVAR was used to assimilate the slant wet delay amounts as initial values. Forecasts up to one hour ahead were made every 1 hour from 10 JST to 14 JST on the 12th (4 forecasts in total). The data assimilation experiment compared forecast accuracy among three types: 1) a control experiment without data assimilation (CNTL), 2) an experiment assimilating zenith wet delay (ZWD) using only satellites within a zenith angle of 15 degrees, and 3) an experiment assimilating slant wet delay (SWD. In the ZWD experiment, only 136 PATHs were usable, while in the SWD, approximately 1200 PATHs became available, offering around ten times more observations. The Fractional Skill Score (FSS) was used as a metric for verifying forecast accuracy. Averaging across 4 forecast experiments, the accuracy one hour ahead was about 0.23 for the CNTL experiment, approximately 0.31 for the ZWD experiment, and about 0.39 for the SWD experiment, demonstrating that assimilating water vapor improves forecasts and that using slant directions enhances accuracy.