The Advances in Targeted Observations for Tropical Cyclone Prediction Based on Conditional Nonlinear Optimal Perturbation (CNOP) Method - Data Assimilation for Atmospheric, Oceanic and Hydrologic Applications - page 587

Table 24.5 Track forecast errors (km) without dropsondes from 24-72 h for all seven cases. The
second row represents the initial time for each typhoon case (all in 2009, e.g., '082012' represents
1200UTC 20 August). The last row indicates the moving direction during this period
Vamco

Mujigae Koppu Choi-Wan Ketsana Mirinae Nida

Initial time
(h)

082012 091000 091306 091700 092618 102918 112806

24 251.3 110 211.6 88.7 168.3 11 125.4

30 305.0 132.9 171.1 85.9 155.6 59.2 213.6

36 323.5 128.8 222.7 125.4 155.6 188.3 287.9

42 365.7 188.3 204.6 192.1 189.9 267.9 282.4

48 487.2 157.5 233.9 275.7 194.9 216.7 412.3

54 364.2 39.7 266.1 268.5 184.4 153.2 493.0

60 261.5 59.2 297.2 260.8 165.4 226.8 376.3

66 354.7 144.7 537.7 49.2 125.1

72 199.2 101.4

Moving
direction

Northward Westward Northwest-

ward

Recurved Westward Westward Stagnant

10 1 of the analysis. The
simulated additional dropsonde data included horizontal wind speed, horizontal
wind direction, and temperature at 850, 500 and 200 hPa. The 3D-Var assimilation
system of MM5 is used to assimilate the additional dropsonde data to produce an
analysis at 24 h, which can be run to predict the locations of typhoon centres in
the following 48 h (from 24 to 72 h). The differences between these typhoon centre
positions and the nature run are defined as the typhoon track forecast errors with
dropsondes. Difference between these errors and those without dropsondes are used
to indicate the influence of CNOPs sensitive areas on typhoon track forecasts.
Seven typhoon events (with large track forecast errors, Table 24.5 ) originated in
the western North Pacific during the 2009 season have been selected for analysis. It
is found that CNOP sensitive areas forms (half) an annulus around the typhoon
centres at targeting time for most of the typhoon cases (five of seven); and SV
sensitive areas showed a maximum at the rear left quadrant with respect to the
storm motion, approximately 500 km from the centre of the storms, also in five
of seven cases (Fig. 24.16 , typhoon Mirinae for example). Then dropping sites have
been selected: the distance between adjacent sites is appropriate (about 150 km),
and the total number of sites is the same in the CNOP and SV sensitive areas. See
Fig. 24.16 for example. The track errors that with and without dropsondes have
been compared. It is found that a varying degree of improvement in typhoon track
forecasts for six of the seven cases, after assimilating simulated dropsonde data
obtained for the sensitive regions (Fig. 24.17 ). Moreover, the improvements are not
only obtained for the optimization period, for calculating CNOPs and SVs, but also
for the subsequent 24 h. During the period 24-72 h, the deployment of dropsondes
according to CNOPs sensitivity could reduce track forecast errors by 13-46 %, and
by 14-25 % for SVs sensitivity.

randomly produced observation errors with the order of