Post on 13-Dec-2015
transcript
1 00/XXXX © Crown copyright
Stratospheric Data Assimilation for NWP
Richard Swinbank
with thanks to:
David Jackson, Mike Keil, Andrew Bushell, Hazel Thornton
and Rick Rawlins
COST-723 Workshop, March 2004
2 00/XXXX © Crown copyright
Stratospheric Data Assimilation for NWP
Variational assimilation: 3D-VAR
New Dynamics
Ozone and water vapour assimilation
Future plans
3 00/XXXX © Crown copyright
Stratospheric Analysis – History
“SSU Analysis” - 1978– Original stratospheric analysis, based on gridded retrievals of
thickness; T and winds derived Analysis Correction Scheme - 1991
– First Met Office stratospheric data assimilation system; asynoptic, repeated insertion
Variational Assimilation – 2000– 3D-VAR assimilation; 6 hour cycle
New Dynamics – 2003– Semi-Lagrangian Dynamics, on height grid
4D-VAR – planned
4 00/XXXX © Crown copyright
3D-VAR for the stratosphere
Direct assimilation of ATOVS & TOVS radiances– radiance bias correction
Background error covariances using “NMC method”– use rotated vertical modes in stratosphere
Prototype for future extended global forecast system, spanning stratosphere
– 40-level model, based on the then-current global 30-L model
– most testing done at medium resolution (0.83°x1.25°), rather than usual stratospheric low resolution (2.5°x3.75°)
7 00/XXXX © Crown copyright
New Dynamics Old Dynamics
Semi-Lagrangian
Semi-implicit (predictor-corrector)
Arakawa C-grid
Height based: hybrid terrain-following grid
Charney-Phillips
Full 3D Helmholtz solver
Explicit Heun
Split-explicit (2 time-level)
Arakawa B-grid
Pressure based: hybrid sigma-pressure grid
Lorenz
Reference state profile
11 00/XXXX © Crown copyright
New Dynamics Configurations
38-level, N216 (0.55o x 0.83o) Top at 39km
Operational (NWP) in August 2002
50-level, N48 (2.5o x 3.75o) Methane oxidation and spectral GWD
Top at 64 km
Operational (NWP) in October 2003
Positive benefit on forecast and analysis skill
Current ND levels
17 00/XXXX © Crown copyright
Ozone Assimilation
Potential benefits for NWP
– Improved radiance assimilation (HIRS, AIRS, IASI)
– Improved radiative heating rates
– Possible impact on UTLS wind fields
– Improved forecasts of surface UV
Exploitation of research satellite data (especially Envisat data, through ASSET)
18 00/XXXX © Crown copyright
Ozone assimilation - status
Initial trial completed with New Dynamics, using HIRS-9 and SBUV
Some bugs remain with Envisat version of suite Further analysis and testing needed:
– Test with Cariolle ozone, cold tracer
– Test with spectral GWD
– Migrate to RTTOV-7 forward model
– Use NMC method for B
– Feedback ozone on radiation
19 00/XXXX © Crown copyright
Stratospheric Humidity Assimilation
Previously untried
Ill-conditioned vertical transform of background error matrix B
Excessive stratospheric increments (due to spurious vertical correlations)
Run 3D-VAR experiments for 1 cycle
VAR humidity tests6 hr cycle – humidity data only in troposphere
q (bg) q (an)
q inc T inc
21 00/XXXX © Crown copyright
Proposed Solutions
Change vertical weighting to improve conditioning of B
Scale back variances, and reduce vertical correlations
New humidity variable (after Dee and da Silva, 2003)
Impact of Dee and DaSilva scheme
q(analysis) q increment
Default
Dee & daSilva ..
23 00/XXXX © Crown copyright
Future Plans
4D-Variational Assimilation– Very promising results from trials of Global
forecast model
– To be implemented soon
Extend Global Assimilation to span the stratosphere
– Improve assimilation of satellite radiances
– Avoids need for separate operational stratospheric configuration
24 00/XXXX © Crown copyright
Data Assimilation
4D-Var in outline
Run forecast from background field
T-3 T-2 T-1 T+0
T+1
T+2
T+3
T+144
Background field
Analysis Backgroun
d field for next forecast
Change starting field to reduce distance between forecast trajectory and observations (error)
Repeat until satisfactory fit
NWP index
H500 anomaly correlation (N216)
Current and Future(??) levels
28 00/XXXX © Crown copyright
Summary
The operational stratospheric assimilation system has developed rapidly over the past few years (3D-VAR in 2000, New Dynamics in 2003, porting to NEC SX-6)
The system has been extended to assimilate ozone and water vapour – in particular as part of ASSET project
Future developments include adoption of 4D-VAR, and merging global and stratospheric assimilation systems