A Numerical Study on a Mesoscale Convective System over a Subtropical Island with 4D-Var Assimilation of GPS Slant Total Delays

Abstract

 A method to assimilate slant total delays derived from global positioning system (GPS) data using the four-dimensional variational data assimilation technique was developed and applied to a line-shaped, local heavy rainfall event that formed on 19 August 2009 over Okinawa Island, Japan.
 First, to identify the primary factors affecting rainband initiation, we performed impact tests using the Japan Meteorological Agency non-hydrostatic model (JMANHM) with 5-km horizontal grid spacing. Simulations in which the orography of Okinawa was removed successfully reproduced the rainband over southern Okinawa, which showed that the primary factor leading to rainband initiation was land surface heating. However, the timing of rainband initiation in these experiments was delayed, and the rainfall intensities were weaker than those observed.
 To reduce these discrepancies, we first conducted a high-resolution numerical experiment using JMANHM with 2-km horizontal grid spacing (NODA) followed by data assimilation experiments with GPS observations (i.e., GPS zenith total delay (GPS-ZTD), GPS precipitable water vapor (GPS-PWV), and GPS slant total delay (GPS-STD)) at the same resolution. As a result, increasing the horizontal resolution improved the simulation of the rainfall intensity. Generally, compared with NODA, the assimilations of GPS-ZTD and GPS-PWV are known to slightly improve the timing of the subsequent rainband initiation. However, the GPS-STD assimilation significantly improved the water vapor and temperature fields over a wide area and yielded a clearly improved forecast in terms of both rainfall timing and intensity.