Assimilation of Doppler radar and surface observations for the tornado outbreak on 6 May 2012

Presentation Type: 
Sho Yokota (Meteorological Research Institute)
Masaru Kunii (Meteorological Research Institute, Japan Meteorological Agency)
Hiromu Seko (Meteorological Research Institute, Japan Meteorological Agency/JAMSTEC)

A strong tornado with F3 scale was generated on the Kanto Plain, Japan at about 12:30 LT on 6 May 2012. It was one of the strongest tornadoes in Japan and caused serious damage. Besides horizontal wind in rainfall regions that caused the tornado, the lower vortex with this tornado was well captured by the Doppler radar in Meteorological Research Institute (MRI-Radar) located about 15km south of the path of the tornado. In addition, a lot of surface wind and temperature observations were available over the Kanto Plain, including the generation point of tornado. However, these high-density observations have not been assimilated so far for this case. In this study, Doppler wind and reflectivity observed by MRI-radar and surface wind and temperature observed by Automated Meteorological Data Acquisition System (AMeDAS) of Japan Meteorological Agency (JMA) were assimilated with a Nested Local Ensemble Transform Kalman Filter (LETKF) system. To reproduce large scale convergence, hourly observation data used in JMA operational model were assimilated with 6 hour intervals by the Outer-LETKF with the horizontal grid interval of 15 km. In the Inner-LETKF with the horizontal grid interval of 1.875 km, the MRI-radar and AMeDAS data obtained every 10 minutes were additionally assimilated with 1 hour intervals. To show the impacts of these data on the tornado outbreak, experiments without MRI-radar or AMeDAS data were also performed. After the data assimilation experiments, the downscaling ensemble experiments were carried out by using the ensemble analyses and perturbations at 11:00 LT on 6 May 2012 as initial conditions. In the downscaling ensemble experiments, strong vortices were generated, and the assimilation of both MRI-radar and AMeDAS data was efficient to improve the position of the tornado.