Every-10-Minute Refresh of Precipitation and Flood Risk Predictions by Assimilating Himawari-8 All-Sky Satellite Radiances

In July 2015, Japan Meteorological Agency started full operations of “Himawari-8”, a third generation geostationary satellite. Himawari-8 is capable of every-10-minute full disk observation similar to GOES-16 and enables to refresh precipitation and flood predictions as frequently as every 10 minutes. This has a potential advantage in capturing the flood risk associated with a sudden torrential rainfall at an earlier stage. The present study aims to demonstrate the advantage of every-10-minute updates of precipitation and flood risk predictions by assimilating all-sky Himawari-8 infrared (IR) radiances. To do so, we use an advanced regional data assimilation system known as the SCALE-LETKF, composed of a regional numerical weather prediction (NWP) model (SCALE-RM) developed in RIKEN, Japan and the Local Ensemble Transform Kalman Filter (LETKF). We focus on a major disaster case in Japan known as September 2015 Kanto-Tohoku heavy rainfall in which a meridional precipitation band associated with a tropical cyclone induced a record-breaking rainfall and eventually caused a collapse of a Kinu River levee. By assimilating a moisture sensitive IR band (band 9, 6.9 µm) of Himawari-8 every 10 minutes into a 6-km mesh SCALE-LETKF, the heavy precipitation forecasts are greatly improved. This helps improve river discharge forecasts by a rainfall-runoff model with longer lead times.