Development of a Two-way Nested LETKF System for Cloud-resolving Model - Data Assimilation for Atmospheric, Oceanic and Hydrologic Applications - page 493

20.5.1 Impact of GPS Water Vapor Data

Due to frequent occurrence of earthquakes, movements of clusters are monitored
by more than 1,200 GPS receivers that have been deployed by the Geospatial
Information Authority of Japan. Because signals of radio waves transmitted from
GPS satellites are delayed by water vapor in the atmosphere, the delays of signals
are estimated as well as GPS receivers' positions. In this study, PWV and SWV
that were estimated from the delays ( Shoji et al. 2004 ), which have information
of water vapor, were used as assimilation data. PWV and SWV are the integrated
value of water vapor in the column or along the paths from GPS satellites to GPS
receivers. For this study, we produced intermediate profiles of relative humidity
from observations and statistical data from outputs of the ensemble forecast, and
these profiles were assimilated by LETKF. In the estimation of intermediate profiles,
the following two assumptions were used; (1) Differences between intermediate
profile and first guess are proportional to the spread of relative humidity. Due
to position errors of rainfall regions and large dispersion of relative humidity
distributions caused by small ensemble size, area-mean relative humidity profile
and area-maximum spread profiles of relative humidity within the areas from
18 km from GPS receivers were used as the first guess and spread profiles of
relative humidity (detailed procedures were explained in Seko et al. 2011 ). (2)
The intermediate profile is produced at the layers where the correlation among
the ensemble members between relative humidity of each layer and PWV exists
( Fujita et al. 2011 ). In this study, the correlation of 0.3 was used as the threshold.
Namely, relative humidity was increased where the correlation was larger than
0.3, and decreased where the correlation was smaller than -0.3. The assimilation
method of SWV was the same as that of PWV, except for the slant paths and the
small areas that were used in producing the ensemble mean and maximum spread
profiles of relative humidity. Because SWV is water vapor between GPS satellites
and receivers, SWV data provides water vapor values as well as its direction. If large
areas were used in producing the ensemble mean and maximum spread profiles,
the direction would become ambiguous because large areas dilute this information.
To exploit this advantage of SWV, areas used in producing ensemble mean and
maximum spread were reduced from 18 to 3 km.
Figure 20.11 shows the rainfall regions at 17 JST that were obtained by assim-
ilation of PWV and SWV data. In addition to convectional data, PWV and SWV
data from 9 to 15 JST were assimilated in the Inner LETKF. When the PWV data
was added, the number of ensemble members in which the rainfall regions were
extending northwestward increased from 4 to 7 (#001-#004, #008, #009 and #011)
(Figs. 20.6 band 20.11 a). Because the assimilation of PWV data modified the water
vapor that was supplied into the rainfall region, the intensities of the rainfall regions
are expected to be improved. In this experiment, the position and intensities of
rainfall regions were improved. When SWV data was added to assimilation data,
the intense rainfall regions were generated at the northwestern side of Sakai City,
where the intense rainfall region was observed (16 JST of Fig. 20.2 b), in most of