1. Development of Data Assimilation Systems for Short-Term Numerical Weather Prediction at JMA
Tadashi Fujita (NPD JMA)
Y. Honda, Y. Ikuta, J. Fukuda, Y. Ishikawa, K. Yoshimoto

2. Contents
1. Meso-scale NWP system (MA: Meso-scale Analysis)
1-1. MA operational system
1-2. recent update
2. Local NWP system (LA: Local Analysis)
2-1. LA trial operation system
2-2. recent developments
3. Summary

3. Contents
1. Meso-scale NWP system (MA: Meso-scale Analysis)
1-1. MA operational system
1-2. recent update
2. Local NWP system (LA: Local Analysis)
2-1. LA trial operation system
2-2. recent developments
3. Summary

4. Meso-scale NWP System
Forecast Model : Meso-Scale Model (MSM) based on JMA Nonhydrostatic Model (JMA-NHM)
Data Assimilation System : Meso-scale Analysis (MA) based on Nonhydrostatic meso 4DVar-system (JNoVA)
Domain
Specifications
Horizontal resolution：5km
Domain: 3600*2880km (721*577 grid points)
Forecast term + 00,06,12,18Z => 15hours + 03,09,15,21Z => 33hours
Forecast model
Meso-Scale Model (MSM)
Initial condition (atmosphere) 　 Meso-scale Analysis (MA)
Boundary condition 20km-GSM (Global Spectral Model)

5. Objectives of Meso-scale NWP System
Disaster Prevention
Prediction of severe weather such as heavy rainfall is one of the main targets for mitigation and reduction of damage to property and loss of life.
Input to short-range precipitation forecast system
Input to storm surge model
Aviation Weather Forecast
Enrichment of the weather information for aviation safety
Terminal Area Forecast (TAF) Guidance and so on.

6. MA operational system MA MSM 00UTC 03UTC 03UTC 06UTC FG (5km JMA-NHM)
Obs. - FG
Obs. - FG
JNoVA 4DVar
(inner model 15km JMA-NHM)
JNoVA 4DVar (inner model 15km JMA-NHM)
Analysis increment
Analysis increment MA
Outer model 5km JMA-NHM
outer model 5km JMA-NHM
MSM
MSM (5km JMA-NHM)
33h forecast
15h forecast

7. MA Coverage Maps of Observation Data7

8. Coverage Maps of Observation Data
Direct assimilation of satellite radiance data 8

9. Score of MSM Precipitation Forecast
Verification Grid : 20km Square
Verified Element: 1mm/3hr
Verification Period :From Mar to Sep. 2011
Threat Score
Radar reflectivity
20km GSM
satellite radiance temperature
GPS
Nonhydro model
dx=10km =>5km
Major revision of physical processes
Nonhydro 4DVar
4DVar
Improvement of convective scheme 9 9 9

10. Contents
1. Meso-scale NWP system (MA: Meso-scale Analysis)
1-1. MA operational system
1-2. recent update
2. Local NWP system (LA: Local Analysis)
2-1. LA trial operation system
2-2. recent developments
3. Summary

11. Use of 3D radar reflectivity data
(started 9 Jun. 2011)
Ze from Radar simulator
First Guess (MSM)
Ze obs.
Assimilation of RH data retrieved from3D radar reflectivity
=> 　Improvement of humidity and precipitation forecast of MSM
RH retrieval algorithm
retrieved RH
retrieved RH
retrieved RH MA
inner model (15km)
MSM (5km)
Outer model (5km)
MSM 3h accumulated precipitation forecast
26 Jul UTC
(cf. Meteo France method) 11

12. Contents
1. Meso-scale NWP system (MA: Meso-scale Analysis)
1-1. MA operational system
1-2. recent update
2. Local NWP system (LA: Local Analysis)
2-1. LA trial operation system
2-2. recent developments
3. Summary

13. Local NWP System
Forecast Model : Local Forecast Model (LFM) JMA Nonhydrostatic Model (JMA-NHM)
Data Assimilation System : Local Analysis (LA) JNoVA 3DVar
Trial operation started in Nov. 2010
operation planned in 2012
Specifications
Objectives
Producing sophisticated disaster prevention and aviation weather information with high resolution NWP
Horizontal resolution：2km
Forecast term + 9hours Forecast model
Local Forecast Model (LFM)
Initial condition (atmosphere) 　 Local Analysis (LA)
Boundary condition 5km-MSM
domain used in trial operation
LFM (2km 551x801)
LA (5km 441x501)

14. LA trial operation system
MSM (in operation)
FT=3
Rapid update cycle (RUC) 3DVAR
3DVAR (5km)
3DVAR
3DVAR
3DVAR
LF1 (5km)
LF1
LF1
LFM (2km) LA
MSM (in operation)
FT=3
FT=-3
FT=-2
FT=-1
FT=0
Boundary Condition
First Guess
hydrometeors
Analysis
LF1
JMA-NHM 1h forecast, dx=5km

15. LA Coverage Maps of Observation Data
Surface stations
(temperature and wind)
Wind Profiler
(horizontal wind)
Doppler radar
(radial velocity)
Aviation(temperature
and horizontal wind)
Ground-based GPS
(total column water vapor)

16. LFM precipitation forecast
precipitation related to heated land in the afternoon (16 Aug UTC 1h precipitation)
Observation
LFM (FT=3)

17. Contents
1. Meso-scale NWP system (MA: Meso-scale Analysis)
1-1. MA operational system
1-2. recent update
2. Local NWP system (LA: Local Analysis)
2-1. LA system in trial operation
2-2. recent development
3. Summary

18. (i) Use of radar reflectivity observation
Simulate radar reflectivity from LF1 (JMA-NHM forecast) => estimate RH from reflectivity => assimilate RH in 3DVAR
RH - reflectivity Database
3DVAR RH
retrieval Ze Ze
Rain, snow, graupel
Radar simulator
LF1
LF1
Radar obs.

19. 3h accumulated precipitation（FT=3）
(i) Use of radar reflectivity observation
Control
Test
Observation
3h accumulated precipitation（FT=3）
FT=0 Total column water vapor (Test-Control)

20. (ii) Vertical Coordinate of Control Variable
Control(=Trial Operation) : z*-coordinate
Influence of topography remains strong up to high altitudes
Test : New coordinate
follow terrain near the surface => rapidly shift to z-coordinate aloft
z*-coordinate
CV new coordinate
model top
z-coordinate
z-coordinate
Slowly shift to
z-coordinate
Rapidly shift to
z-coordinate
ground

21. (ii) Vertical Coordinate of Control Variable
Vertical Cross Section of T increment
Control
Test
Reasonably limits the influence of topography within the lower troposphere.

22. (iii) Extension of Control Variables
Control: ground potential temperature is fixed
⇒excessive temperature increment in the lower troposphere
Test : extend the control variable to include ground potential temperature
⇒ Analyze ground PT to mitigate excessive increment
excessive increment
ground
1.5m
20m 0m
surface
the lowest model level PT
Obs
1.5m
20m 0m
Analyze ground PT
fixed

23. (iii) Extension of Control Variables
Vertical cross section of temperature analysis increment
Control
Test
Mitigate excessive temperature increment in the lower troposphere

24. (iv) Incremental Analysis Updates
Gradually add 3DVar increment in the assimilation window
=> enhance balance of the analysis
MSM
Obs.
Obs.
Obs.
Obs.
3DVar
3DVar
3DVar
3DVar
30min.
Gradually add increment
LFM
(cf. Bloom et al. 1996, Clayton 2003, Lee et al. 2006, etc.)

25. (iv) Incremental Analysis Updates
Gradually add 3DVar increment in the assimilation window
=> enhance balance of the analysis
Actual implementation in test experiment
MSM
Obs.
Obs.
Obs.
Obs.
3DVar
3DVar
3DVar
3DVar
30min.
5km JMA-NHM

26. (iv) Incremental Analysis Updates (Test with 5km forecast)
Domain averaged Ps tendency
Qc summed over (limited) domain
update of B.C.
Control
Test
Control
Test
Rapid update cycle
FT=6h
5km forecast
FT=0
5km forecast
FT=-3h
FT=0
FT=6h

27. (v) Terrain-Adjusted Background Error Correlation
Terrain between grid points is used to modify horizontal background error correlation (steep => damp fast)
Implemented using coordinate transformation + recursive filter
Test by single surface T observation
(T increment on the lowest model level)
Control
Test

28. Summary
JMA operates Meso-scale NWP system aimed at disaster prevention and aviation weather information services.
Steady improvement of MSM forecast has been attained from various improvements of the system, including recent introduction of radar reflectivity data (retrieved RH) in MA.
Trial operation of Local NWP system is currently underway, toward the operational run scheduled in 2012.
Various development of LA is underway to improve the system.
introduction of new observation, including radar reflectivity data
new CV vertical coordinate
ground PT analysis
IAU
terrain-adjusted background error correlation