MM5 3DVAR Impact Studies

1
MM5 3DVAR Impact Studies on Weather Prediction
for COSMIC   1Ching-Yuang Huang, 2Ying-Hwa Kuo
and 2Wei Huang 1Department of Atmospheric
Sciences, National Central University 2National
Center for Atmospheric Sciences, USA    

• MM5 3DVAR system developed in NCAR has been
  implemented in NCU and is used to investigate the
  influence of ingested data on short-term weather
  forecast.
• Three weather events are investigated,
  including Mei-Yu front on 06/1998 Bilis
  super-typhoon on 08/2000 Nari typhoon on
  09/2001.  

2

• Data assimilation has recently recognized as a
  mean of providing better consistent initial
  conditions for numerical weather prediction in
  the meteorological community.
• One of the most attractive and effective
  methods is the utilization of estimation theory
  based on variational analyses to minimize the
  bias of analysis data. This treatment through the
  so-called Kalman filter may provide theoretically
  best or optimal solution for the analyzed
  system (Zou et al., 1997).
• A full set of data in all assimilation time
  window has the beneficial impact resulted from
  the strong constraint upon the model integrated
  state with both physics and dynamics and is well
  known as 4DVAR, which, however, is very
  time-consuming due to the adjoint nature in
  iteratively searching for the optimal solution.
• The 3DVAR which utilizes both the observations
  and analyses at the current time (i.e., the
  initial time) and greatly simplifies the
  filtering process subject to adjoint operators
  for observations (Vandenberghe and Kuo, 1999).
• Another advantage of the 3DVAR, as compared to
  4DVAR, is the mathematic configuration on
  multi-domains which facilitate nested simulations
  for multiple-scale phenomena with conventional
  and unconventional data.
• We utilize MM5 3DVAR to incorporate sounding
  observations with the diagonal error covariance
  matrix provided by NCEP for the operation
  Spectral Statistical Interpolation 3DVAR system
  (Parrish and Derber, 1992).

3
3DVAR Analysis Define a cost function as
Here x Analysis variables(N-dimensions),
xb Background variables(N-dimensions),
yobsObservations(M-dimensions), B
Background error covariance matrix(N N), O
Observation error covariance matrix(M M), R
Transformation of analysis variables to
observation variables (may be
nonlinear) , N Freedom of analysis, M
Number of observations.
,
J Jb Jo Jc
constraint on x
4

• All the cases have been simulated by MM5 version
  3.4 with ice/graupel physics in the finest domain
  and Grells scheme for cumulus parameterization
  in outer domains and the Blackadar PBL
  parameterization in all the domains.
• An objective analysis module (Little_R) has also
  been used to ingest the sounding observations
  prior to 3DVAR.
• The domain setups
• Mei-Yu front case Two domains
• 45 and 15 km resolutions, 72 h forecast.
• Bilis typhoon case Three domains
• 45, 15 and 5 km resolutions, 60 h forecast.
• Nari typhoon case Three domains
• 45, 15 and 5 km resolutions, 72 h forecast.

5
a
b
d
c
6
a
b
d
c
7
a
b
b
c
d
Fig. (a) Cloud water contents (g kg-1) at 500 mb
at 1800 UTC 03 June 1998 for the no-3DVAR run
without LITTLE_R, (b) as in (a) but at 2100 UTC,
(c) as in (a) but for the 3DVAR run without
LITTLE_R, and (d) as in (c) but at 2100 UTC.
8
a
Fig. IR images (a) at 2333Z/03/06/1998 and (b)
0033Z/04/06/1998.
9
a
b
b
a
c
d
,
10
a
b
11
a
b
Fig. The observed accumulated rainfall amount
for Bilis typhoon (a) during 1200Z1500Z/22/08
(88 mm) and (b) 1500Z1800Z/22/08 (172 mm).
12
a
b
d
c
13
a
b
c
d
14
a
b
Fig. 9. The differences between the initial
fields for the Bilis typhoon case at 0000 UTC 21
August 2000 for the runs with and without 3DVAR.
Shown are perturbation pressure (mb) and
horizontal wind (m s-1) at s 0.995 in (a)
domain 1 and (b) domain 2.
15
a
b
d
c
f
e
g
h
Fig. 9. Infrared imagery at every 6 h from
0000Z/16/09/2001 to 0018Z/17/09/2001.
16
b
a
c
d
Fig. 10. The observed accumulated rainfall
amount for Nari typhoon (a) during
2100Z/16/090000Z/17/09 (230 mm), (b)
0000Z0300Z/17/09 (124 mm), (c) 0300Z0600Z/17/09
(124 mm) and (d) 0600Z0900Z/17/09 (236 mm).
17
a
b
c
d
18
a
b
c
d
19
a
b
c
d
20
a
b
c
d
21
a
b
22
a
b
c
d
Fig. The differences between the initial fields
of Nari typhoon initialized at 0000Z/16/09/2001wit
hout 3D-VAR and with 3D-VAR for perturbation
pressure (mb) and horizontal wind (m s-1). (a)
the outer domain 1 at s 0.995, (b) the domain 2
at s 0.995, (c) the domain 2 at s 0.525 and
(d) the domain 3 at s 0.995.
23
Vmax 70 m/s Rmax 50 km
24
Typhoon Nari (2001)Case Simulations with a
Bogus Vortex
25
a
b
c
d
Fig. The simulated sea-level pressure (mb) and
near-surface horizontal wind (m/s) in
domain 3 for the Nari case V30R50 run in Sep
2001. (a) 12h (b) 24h (c) 36h (d) 48h
after the initial time .
26
b
a
c
d
Fig. The simulated sea-level pressure (mb) and
near-surface horizontal wind (m/s) in
domain 3 for the Nari case V30R50n run in Sep
2001. (a) 12h (b) 24h (c) 36h (d) 48h
after the initial time .
27
a
b
c
d
Fig. The simulated sea-level pressure (mb) and
near-surface horizontal wind (m/s) in
domain 3 for the Nari case V40R50 run in Sep
2001. (a) 12h (b) 24h (c) 36h (d) 48h
after the initial time .
28
b
a
d
c
Fi . The simulated sea-level pressure (mb) and
near-surface horizontal wind (m/s) in
domain 3 for the Nari case V40R50n run in Sep
2001. (a) 12h (b) 24h (c) 36h (d) 48h
after the initial time .
29
a
b
c
d
Fig. The simulated sea-level pressure (mb) and
near-surface horizontal wind (m/s) in
domain 3 for the Nari case V40R100 run in Sep
2001. (a) 12h (b) 24h (c) 36h (d) 48h
after the initial time .
30

• CONCLUSIONS
• MM5 3DVAR system in NCAR has been used to
  investigate the influence of ingested data on
  short-term weather forecast. Three weather
  events are investigated, which include a Mei-Yu
  front on 06/1998 and Bilis super-typhoon on
  08/2000 and Nari typhoon on 09/2001.
• With 3DVAR, the simulated low pressure system
  extending off northeast of Taiwan for the Mei-Yu
  front event is stronger than that without 3DVAR.
  The 24-h predicted heavy rainfall just off
  southern Taiwan and the propagation of the
  cut-off low center are also closer to the
  observed in the 3DVAR run.
• In the second case of Bilis typhoon, both the
  runs with and without 3DVAR show a northward
  track bias just upstream of eastern Taiwan as the
  initialization starts earlier on 0000Z/21/08/2000
  about 39 h prior to landfall. The 3DVAR run still
  gives slightly better rainfall distribution as
  compared to the one without 3DVAR.
• The simulated Nari typhoon event starting from
  0000Z/15/09/2001 shows that cloud convection is
  significantly enhanced as the vortex core is
  pushed to confront with the leading edge of
  Central Mountain Range (CMR) and is slowed down
  to rotates about the left side of CMR.
• The combination of compression and stagnation of
  the embedded convective system along CMR may
  explain the very large rainfall intensity in
  northwestern Taiwan. The feature of the observed
  large rainfall in the southwestern region off CMR
  is also captured but its intensity is
  significantly underpredicted due to lagging or
  weakening of the vortex center at later times in
  the runs with or without 3DVAR.
• With the 3DVAR, the simulated Nari track is
  considerably improved, but the associated
  rainfall patterns remain similar.
• Several 3DVAR initializations with a bogus vortex
  center are attempted for the improvement of the
  rainfall prediction, but the forecast results are
  not improved in terms of rainfall and track
  prediction.
•  

31
COSMIC Meteorological Research in
Taiwan ProjectCOSMIC Data Verification and its
Application on Short-Term Weather and Climate
Prediction PI Huang Ching-Yuang
(???) Co-PIs Liou Yuei-An (???)?
Wu Chun-Chieh (???)?
Chang Jen-Cheng (???)
32
Taiwan-COSMIC Meteorological Research

33
Project Missions

• The goals of the Taiwan-COSMIC Integrated
  Meteorological Research Project are to promote
  and pursue the COSMIC missions on academic
  research, practical needs and facilitate and
  enhance the technical collaborations between
  different programs and to develop the integrated
  research programs.
• One major goal of the GENERAL project is to
  CONDUCT COSMIC real-time numerical weather
  prediction (NWP) experiment to ASSESS GPS/MET
  data impact on Taiwan local weather prediction.

34

• The science team will collaborate to ingest
  COSMIC data into CWB global spectral model and
  CWB MM5 with 3DVAR assimilation and MM5/Adjoint
  model with 4D data assimilations to test the
  readiness of these models and investigate the
  performance of the entire model/operation
  assimilation system.
• The results from the Taiwan-COSMIC real-time NWP
  experiment would provide complete reports on the
  evaluation of GPS/MET data on improvement of
  short-term weather/climate prediction in Taiwan
  area and give useful references for any continued
  satellite program in Taiwan related to COSMIC
  meteorology.

35

• CAA and CWB have agreed to provide helps in data
  ingestion and collaborate to develop and modify
  the operational MM5 3DVAR system in Taiwan.
• We seek for international collaborations to
  achieve the above missions and advance our
  scientific works.
• By 2005, the data assimilation system in Taiwan
  should be ready with TACC.
• Data considered for ingestion include
• GPS/MET data set (1995-1997)
• CHAMP data (for test)
• SAC-C data (for test)
• SSM/I (ok)
• QuickScat (ok)
• Taiwan GPS Ground-based PW (ok)
• COSMIC GPS/MET data (2005 )

36
Project 1

• Integrate COSMIC and ground-based GPS network to
  increase the accuracy of occultation retrieval.
  The scheme of tomography will be developed to
  incorporate ground-based GPS measurements into
  the occultations for improving retrievals of
  atmospheric profiles from the occultations.
• The error sources of the retrievals from
  occultations will be investigated. A
  research-oriented data validation center will be
  set up to promptly support the other scientists
  of our GPS meteorology team.

37
Project 2

• Perform numerical simulation and the impact study
  using MM5 4D-VAR data assimilation system. The
  objective of this research is to improve the
  models simulation capability, the sensitivity
  areas, and the impact of special observational
  data (such as PW and GPS/MET) on typhoon and
  other weather simulation, through the advanced
  MM5 adjoint data assimilation system.

38
Project 3

• Complete the implementations of the so-called GPS
  ray-tracing observation operator and its adjoint
  system base on the CWBs global model structure.
  They will be applied to set up the procedures to
  check on the retrieved data using currently
  available GPS/MET data from the MicroLab II.
• The tools then-established will next be used to
  evaluate the accuracy and the impact of the
  COSMICs GPS/MET observations in the later 2
  years of the project after the launch.

39
Project 4

• Promote the cooperation of research direction
  under the Integrated Research Project.
• Implement MM5 3DVAR data assimilation system for
  ingestion of the GPS/MET data to improve weather
  forecast on real-time mode.

40
The GPS Occultation Geometry
Forward Propagation
Abel Inversion
41
Obtaining Temperature and Pressure From
Refractivity
42
CDAAC Data Volume
43
CDAAC Function