Title: Typhoon Predictions with GPS Radio Occultation Data Assimilations Using WRF 3DVAR with Local and Non
1Typhoon Predictions with GPS Radio Occultation
Data Assimilations Using WRF 3DVAR with Local
and Nonlocal Operators
- Shu-Ya Chen1, Ching-Yuang Huang1, Ying-Hwa
Kuo2,3, - Yong-Run Guo3 and Sergey Sokolovskiy2
- 1Department of Atmospheric Sciences, National
Central University, Jhongli, Taiwan - 2University Corporation for Atmospheric Research,
Boulder, Colorado, USA - 3National Center for Atmospheric Research,
Boulder, Colorado, USA
2Introduction
- Global Positioning System (GPS) radio occultation
(RO) measurements have several advantages, e.g.,
high vertical resolution, no calibration need, no
influence by cloud or rainfall, and global
coverage. - GPS RO refractivity is an intermediate product
after taking an Abel transform of bending angles
defined for each incident ray tangent to its
perigee point. - Currently, assimilation with the Abel-retrieved
refractivity obtained from GPS RO data has
assumed local measurement at the perigee point
where the GPS ray is closest to the earth (e.g.
Huang et al. 2005).
Local (refractivity)
where Pw is water vapor pressure, T air
temperature, and P the pressure of the
atmosphere.
3- However, the Abel-retrieved refractivity
accounts for an integrated amount of refractivity
along the total path of the ray in a
spherically-symmetric atmosphere. But, bending
angle assimilation is not feasible for regional
models and is not cost effective due to ray
tracing. - In order to take the effect of the integration
into account, Sokolovskiy et al. (2005) suggest
use of excess phase for assimilation, which is an
integrated amount of the refractivity along the
ray.
Nonlocal (excess phase)
- In this study, we extend the previous work to
assimilate the excess phase to account for the
nonlocal effect along the ray. In this
preliminary study, we develop a nonlocal operator
and implant it into the WRFVAR version 2.1. We
will compare the results with assimilation by
both local and nonlocal operators in simulation
of Typhoon Kaemi (2006).
4Nonlocal operator
- Interpolate both observation and model
refractivity into the mean heights of the model. - Integrate the observed refractivity along the ray
path straightline. - Consider blocking effect when the ray is
interrupted by model boundaries and mountains.
5One month statistic obs. error by
Hollingsworth-Lonngberg method (2003/08/152003/09
/15)
horizontal resolution 45 km grids
22212830 data CHAMP 314 points in the domain
height (km)
6Typhoon Kaemi (initial time2006/07/23 00 UTC)
- Domain 1 15115131 with resolution 45 km
- Domain 2 15115131 with resolution 15 km
- Domain 3 15115131 with resolution 5 km
- Initial field AVN analysis
- OBS Data FORMOSAT-3/COSMIC
- Model WRF v2.1.2
- Integration 72 hours
NONE no assimilation REF assimilated
refractivity by local operator EPH assimilated
excess phase by nonlocal operator
7T q N
EPH
484 mb int.0.05 ? int. 0.1 g kg-1
int. 0.5 N-units
T q N
REF
8EPH
(AN-BN) / BN 100
Interval 0.3
2259
0029
0030
0050
0203
0211
0235
9REF
2259
0029
0030
0050
0203
0211
0235
10Cold Start Experiments Tracks and Intensity
Best track (dots) NONE (EXP0) REF (EXP1) EPH
(EXP2)
11Observational rainfall
Int. 10 mm
Int. 30 mm
Int. 30 mm
12NONE
757
Max. 126.1 mm
294.9 mm 1024 mm
13REF
900
Max. 114.8 mm 271.8 mm
908.6 mm
14EPH
350
Max. 157.4 mm 241.7 mm
912 mm
15differences of wet refractivity (EPH-NONE)
The differences of wet refractivity (N-unit)
between EPH and NONE at 0.5 km height at (a) 48 h
and (b) 72 h. The horizontal wind vectors (m s-1)
for EPH are also overlapped.
16Cycling with BogusEPH assimilation
- TC Bogus pressure and wind field
- EPH GPS excess phase
All assimilations for 3 domains
17Observation distribution
2006/07/22 00 UTC
2006/07/22 06 UTC
2006/07/22 18 UTC
2006/07/23 00 UTC
2006/07/22 12 UTC
18Obs. number
19Cycling experiments Tracks and Intensity
Best track (dots) BogusEPH (EXP3) Bogus
(EXP4) EPH (EXP5)
20Precipitation (second day)
OBS Bogus
BogusEPH
21Conclusion
- The analysis increments in WRF 3DVAR produced by
both nonlocal operator and local operator are
quite similar in magnitudes and spatial
distributions. However, the increments of the
nonlocal operator are of broader vertical scale
and show elongation along the rays direction,
when compared with those of local observation
operator. - For cold start experiments, the use of nonlocal
observation operator produces better
precipitation forecast than local observation
operator. - The best forecast of typhoon track and intensity
is obtained with (i) cycling, (ii) typhoon bogus
and (iii) excess phase assimilation. - The assimilation of COSMIC GPS RO data with
nonlocal observation improves precipitation
forecast when compared with cycling experiments
with typhoon bogus.
22The end