Development of a Two-way Nested LETKF System for Cloud-resolving Model - Data Assimilation for Atmospheric, Oceanic and Hydrologic Applications - page 491

a b

(mm)

(m/s)

Fig. 20.9 Ensemble mean distributions of rainfall ( shaded regions ) and horizontal wind at the
height of 20 m ( vectors ) reproduced by the Inner LETKF at 17 JST obtained from ( a ) the initial
conditions of the Outer LETKF's forecast (CNTL) and ( b ) the analyzed fields of the no-cost
smoother. An arrow in ( b ) indicates the rainfall region that was not generated in CNTL and
generated in the no-cost smoother experiments

smoother experiments were close to the boundary of Inner LETKF, it is deduced
that the observation data just outside of the Inner LETKF's domain modified the
analysis of the Inner LETKF through the initial seeds and boundary conditions of
the Inner LETKF. This result indicates that the no-cost smoother can improve the
rainfall distribution of the Inner LETKF. However, the whole observation data in
the assimilation window period (6 h from the initial time of assimilation in this
experiment) needs to be waited for if the no-cost smoother is used. If real-time
analysis is required, using the initial seeds and boundary conditions produced from
the Outer LETKF's forecast might be a more realistic approach.
Next, the initial seeds of the experiments using the no-cost smoother are
explained. When the initial seeds are produced by the interpolation of the analysis
of the Outer LETKF, the small-scale disturbances that cannot be resolved by the
grid interval of the Outer LETKF are not included in the initial seeds. To reproduce
local heavy rainfalls, small-scale disturbances should be included in the initial seeds
because they are expected to affect the generation of the convection cells. To show
the impact of small-scale disturbances in the initial seeds, the experiments were
performed with the Inner LETKF of the grid number of by using the

interpolated fields of the Outer LETKF ( L at 9 JST in Fig. 20.4 a) and the last
analyzed fields of the Inner LETKF ( S at 9 JST in Fig. 20.4 a), as the initial seeds at
9 JST (indicated by an open pentagon in Fig. 20.4 a). Figure 20.10 shows the spread
fields of rainfalls at 11 JST, after 2 cycles of the Inner LETKFs (indicated by a
gray pentagon in Fig. 20.4 a). During 2 cycles of the Inner LETKF, the boundary
conditions of the Inner LETKF that were produced by assimilation of conventional
data in the Outer LETKF were used, and the conventional data was assimilated from
0910 JST to 1100 JST. These procedures from 0900 JST to 1100 JST are common to
two experiments. Because the mesoscale convergence was reproduced in both seeds,
the rainfall regions were similar to each other (not shown). However, relatively large

121 121