The Assimilation of Himawari-8 Data in High Resolution Regional Numerical Forecast to Support the Meteorological Applications

The numerical weather forecast is fundamental and crucial to the meteorological application and service. The more accurate and detailed the forecast is, the more precise and skilled the meteorological service goes. The assimilation of the increasing number and type of observation, to get the much more accurate initial field, is one of the effective ways to improve the accuracy of numerical weather prediction. Among the observation, satellite remote sensing is an important data resource. At present, microwave and high infrared hyperspectral sensor onboard polar-orbiting meteorological satellite plays the most key role because of their vertical sounding capability. However it is greatly constrained by the observation frequency. Recently, The Japan next-generation geostationary satellite Himawari-8 was successfully launched on 17 October 2014. The 16 channels Advanced Himawari Imager (AHI) flown on the satellite can produce images with a resolution down to 500m and can provide full disk observations every 10 mins. Specifically, there are three water vapor channels in its channels 7-16 covering from the short- to the thermal-infrared bands to provide with the humidity information of the high, middle and low atmospheric layer. Its predecessor, along with the American GOES and the Chinese FY2, just has only one water vapor channel.

The assimilation of Himawari-8 AHI data, firstly focusing on the three water vapor channels, is addressed to study the effect on the Quantitative Precipitation Forecast (QPF) of our meteorological application and service. The high resolution regional numerical forecast system is composed of data assimilation system GSI and numerical forecast model WRF. The assimilation modules of Himawari-8 AHI data, including the variable definition, data input/output, radiative transfer model, bias correction and quality control, are extended in the GSI framework. The error variance and the bias of AHI measurements, together with the adjustment of bias correction and quality control, are investigated before conducting the data assimilation experiment. The numerical trial and examination are then performed on the first heavy rainfall process occurred in Beijing on 19-20 July 2016, which is comparable to that case in every 40 years on 21-22 July 2012. The comparison between the single assimilation at initial time and multiple assimilations in 12 hours time window is done to reveal the value of high frequency observation. In addition, the comparison between the result of three water vapor channels and that of one channel similar to the Himawari-8 predecessor is scheduled to demonstrate the benefit of vertical sounding enhancement.