JMASV(THORPEX_ISS_0412)

1
Numerical model framework for typhoon prediction
at the Japan Meteorological Agency
The 63th Interdepartmental Hurricane Conference
4 Mar. 2009

• Munehiko Yamaguchi12, Takuya Komori1, Takemasa
  Miyoshi13, Masashi Nagata1 and Tetsuo Nakazawa4
• (myamaguchi_at_rsmas.miami.edu)
• Numerical Prediction Division, Japan
  Meteorological Agency
• University of Miami
• University of Maryland
• Meteorological Research Institute, Japan
  Meteorological Agency

2
Office of Numerical Prediction Division at JMA
3
Present status of typhoon forecasts at JMA
JMA issues forecasts up to 3 days as of 2008. We
plan to extend the forecast range up to 5 days.
4
Progress behind the planned 5 day forecasts
Time series of annual average position errors in
Tropical Cyclone (TC) Track Forecasts by the JMA
Global Spectral Model - Western North
Pacific from 1997 to 2007 (three-year running
mean) -
5
Time line of the upgrade of the systems
Topography of 20kmGSM
Topography of TEPS
20kmGSM JMA Global Spectral Model. TYM Typhoon
Model. WEPS One-Week Ensemble Prediction System.
TEPS Typhoon Ensemble Prediction System.
6
Two NWP systems supporting the forecasts

• 20kmGSM (deterministic track and intensity
  forecast)
• 20km GSM runs 4 times a day (00, 06, 12 and 1800
  UTC) with a forecast range of 90 hours except for
  12UTC where it is 216 hours.
• The data assimilation system is the 4DVAR, which
  has been in operation since 2005, and a typhoon
  bogus technique is used.
• Typhoon Ensemble Prediction System (deterministic
  track forecast for the extended forecast period
  and confidence information on track forecast)
• TEPS uses the lower resolution version of 20kmGSM
  (TL319L60)
• TEPS also runs 4 times a day with a forecast
  range of 132 hours for TCs in the responsibility
  area of RSMC Tokyo Typhoon Center.
• The ensemble size is 11 and singular vectors are
  used to make initial perturbations.

7
Typhoon Bogus Technique
Step1. Create a typhoon structure, considering
the asymmetry, based on TC central position,
central pressure and the radius of 30kt wind,
which are analyzed by forecasters at JMA.
Step2. Pick up points from the created structure
(orange dots) and assimilate them in the 4DVAR
assuming that they are observation data.
TC central position, central pressure and the
radius of 30kt wind analyzed by forecasters at
JMA are reflected into the initial state of
20kmGSM
8
Performance of the two systems

1. TEPS of 2007 (quasi-operation)
2. TEPS of 2008 (operation)
3. 20kmGSM of 2008

9
TEPS provides better deterministic forecasts
Ensemble mean track forecasts statistically have
smaller position errors than those of control
run. The error reduction is about 40 km at 5-day
forecasts, which reduction corresponds to the
gain of half a day lead time.
Verification of track forecasts1 verification
period May to Dec., 2007
Black line Control run Red line Ensemble
mean2 Black dots number of sample
1. The TC strength of L is not included in this
verification 2. Ensemble mean tracks are defined
using more than 1 ensemble member
10
TEPS provides confidence information
Strong relationship between ensemble spread and
position error of ensemble mean track forecasts
Number of sample1 149
1.The TC strength of L is included in this
verification 2. Ensemble mean tracks are defined
using more than 1 ensemble member
Position Errors of Ensemble Mean at 5-day
forecasts (km)
Ensemble spread of TC positions2
(ensemble spread accumulated
from 0 to 120 hours forecasts every 6 hours)
11
Optimization of the probability circle
The ensemble spread of TEPS would allow us to
convey confidence information by optimizing the
size and shape of the probability circle. The
development of an application is now under way.
Initial time 2007.07.29 00UTC
Initial time 2007.09.2 18UTC
Confidence C
Confidence B
Confidence A
Confidence A
12
TEPS in 2008
20kmGSM TEPS Control Member (TL319L60) TEPS
Ensemble Mean (TL319L60)
TEPS Control is much worse than 20km GSM. The
benefit of Ensemble Mean with respect to the
control had gone
13
What is the difference of TEPS between 2007 and
2008 ?
In 2007, the model and data assimilation has the
same horizontal resolution, TL319. In 2008, the
analysis field for TEPS was created by
interpolating the analysis field with a
horizontal resolution of 20km, which might cause
an unbalanced state in the initial field of TEPS.
Miyoshi et al. (2009) showed that the track
forecast of TEPS Control has improved by applying
the 4DVAR to the interpolated analysis field,
which is a TL319L60 resolution. Compared to the
4DVAR for 20kmGSM, the computer resources for the
above 4DVAR is negligible.
14
Intensity forecast by 20km GSM (2008)

• Definition of development stage, maturation stage
  and dissipation stage is based on the differences
  of central pressures from initial time to the
  forecast time of 72 hours
• development stage -10hPa gt ?P
• maturation stage -10hPa lt ?P lt 10hPa
• dissipation stage ?P gt 10hPa

15
Summary

• JMA will extend the forecast range from 3 days
  to 5 days.
• Typhoon EPS will support the extended forecast
  range.
• TEPS will be useful in presenting confidence
  information on track forecasts. (an application
  is under development)
• The deterioration of the control forecasts of
  TEPS in 2008 would be solved by another 4DVAR for
  a TL319L60 resolution.
• For the intensity forecasts by 20kmGSM, there is
  a room for improvement, especially for the
  forecasts of tendency of intensity changes.