Title: Tenth ECMWF Workshop on Meteorological Operational Systems
1Tenth ECMWF Workshop onMeteorological
Operational Systems
- WCMWF, Shinfield Park, Reading, Berks.,
- United Kingdom
- 14-18 November 2005
- Summarized By
- Yuejian Zhu
- Environmental Modeling Center
- NOAA/NWS/NCEP
2Main Sessions
- Use and interpretation of medium and extended
rang forecast guidance - Configurations/future plans
- Performance/evaluation
- Applications (include short/regional and monthly
climate studies) - User feedback
- Operational data management systems
- Meteorological visualization applications
3Configurations
- ECMWF Deterministic/ensemble system
- The new high resolution forecast system
- The new variable resolution ensemble prediction
system - Wave forecasting
- NCEP Recent developments with NAEFS
- The new implemented GEFS (August 2005)
- CMC Reviewing the ensemble prediction system
- UKMet Short-range ensemble prediction system
4ECMWF - The New High Resolution Systemby Martin
Miller
- High Resolution System
- Deterministic 10d-forecast
- TL799 L91 (?t12min)
- 4D-Var Analysis
- 1st minimization TL95 L91 ?t1h
- 2nd minimization TL255L91 ?t1/2h
- EPS TL399 L62 (?t30min)
- Wave Model 0.36
- Operational System
- Deterministic 10d-forecast
- TL511 L60 (?t15min)
- 4D-Var Analysis
- 1st minimization TL95 L60 ?t1h
- 2nd minimization TL159L60 ?t1/2h
- EPS TL255 L40 (?t45min)
- Wave Model 0.50
-
5Increase in Horizontal Resolution
Total of points in TL511 grid 348,528
Total of points in TL799 grid 843,490
6Only 2 gridpoints for Majorca in TL511. Grid40kms
In TL799 Majorca is covered by 6 gridpoints.
Grid25kms
7Vertical Resolution Increase
- The number of vertical levels for analysis and
deterministic model - increased from 60 to 91.
- Largest resolution increase near the tropopause
- Model top raised from 0.1hPa (65km) to 0.01hPa
(80km).
8Vertical Resolution Increase for the Ensemble
Prediction System (EPS)
- For the EPS the vertical resolution increases
from 40 to 62 levels. - Like the 40-level model, the 62-level model has
only a few levels in the stratosphere (top model
level at 5hPa). - In the troposphere (up to about 200hPa), the
distribution of levels in L62 is identical to the
91-level distribution. i.e. the EPS and
deterministic systems have the same vertical
resolution in the troposphere.
9Computational Cost of the Resolution Increases
- The deterministic 10 day forecast at TL799 L91
is about 4.3 times - more expensive to run than at TL511 L60.
- An analysis cycle with the high resolution
system is about 3.8 times - more expensive than with the operational system
Distribution of cost for the different parts of
the model
10Verification of the High Resolution System
11Verification of the High Resolution System
12Verification of the High Resolution System
13Why VAREPS?
- VAREPS aims to increase the value of the current
EPS in two ways - up to fc d7, by providing more skilful
predictions of small-scale, severe events - after fc d7, by extending the range of skilful
products from 10 to 15 days - VAREPS will also provide the first 2-legs of
ECMWF planned seamless ensemble system, which
will be extended initially to one month, and then
to a longer forecast time. - The key idea behind VAREPS is to resolve
small-scales in the forecast up to the forecast
range when resolving them improves the forecast,
but dropping them when their impact is negligible.
14EPS configurations tested with 51-members (CY28R3)
- The performance of ensembles run in the following
four configurations have been compared for 46
cases (21 cases from warm and 25 from cold
seasons). - Average results are based on the comparison of
500 hPa geopotential height (Z500), 850 hPa
temperature and total precipitation (TP)
forecasts. Case studies have also considered
significant wave height and 850hPa wind.
fc-day 1 2 3 4 5 6 7 8 9 10 11 12 13 14
VAR7D6 TL399L40-1800s TL399L40-1800s TL399L40-1800s TL399L40-1800s TL399L40-1800s TL399L40-1800s TL399L40-1800s
VAR7D6 TL255L40-2700s TL255L40-2700s TL255L40-2700s TL255L40-2700s TL255L40-2700s TL255L40-2700s TL255L40-2700s TL255L40-2700s
T399 TL399L40-1800s TL399L40-1800s TL399L40-1800s TL399L40-1800s TL399L40-1800s TL399L40-1800s TL399L40-1800s TL399L40-1800s TL399L40-1800s TL399L40-1800s TL399L40-1800s TL399L40-1800s TL399L40-1800s TL399L40-1800s
T399
T255 (OPE)
T255 (OPE) TL255L40-2700s TL255L40-2700s TL255L40-2700s TL255L40-2700s TL255L40-2700s TL255L40-2700s TL255L40-2700s TL255L40-2700s TL255L40-2700s TL255L40-2700s TL255L40-2700s TL255L40-2700s TL255L40-2700s TL255L40-2700s
15Summary with key conclusions
- Expected average impact of EPS upgrade
- Results based on the comparison of Z500 and total
precipitation predictions (46 cases, 51mem)
indicate that in the 1st week, VAREPS will
deliver gains of up to 12h, and in the 2nd week
it will give users access to skilful
probabilistic forecasts. - Impact of EPS upgrade on severe weather forecasts
- In the 1st week, VAREPS(T399) will deliver more
accurate predictions of intense cyclonic
developments (both in terms of intensity and
position), wind speed, significant wave height
and precipitation. - The future a seamless ensemble system from day 0
to day 32 - The first cases of 3-leg VAREPS have been
completed. The configuration planned to be
implemented in Q1/2006 will (most probably) be - Day 0- 7 TL399L62DT1800s
- Day 6-15 TL255L62DT2700s
- Day 15-32 TL255L62DT2700s coupled with ocean
model
16NCEP Recent ImplementationChanges 1by
Yuejian Zhu
- Extend T126 portion of forecast after 180 hours
(see new configuration) - This change is intended to improve ensemble
support for 5-10 days and week-2 forecast by
providing high resolution (T126) and continue (no
resolution change) forecast - Results
- Increased spread for week-2 forecast
- Improving probabilistic skill beyond 180 hours
17NCEP GLOBAL ENSEMBLE FORECAST SYSTEM
NEW
CURRENT CONFIGURATION MARCH 2003
- Breeding method
- 24 hours breeding cycle
- 4 cycles per day
- 10 m for each cycle
- Ens. Ctl at t00z only
- Total 45 m in 24 hours
- 4 different resolutions
- 16-day forecasts
18NCEP Recent ImplementationChanges - 2
- Initial perturbation (breeding cycle)
- This change is intended to enable for relocation
of perturbed tropical storm. Tuning initial
perturbation size is for reducing spread for
short-range forecast - Results
- Decreased spread for short-rang (1-3) forecast
- Improving forecast skill for first 3 days
- Improving probabilistic forecast skill for short
lead-time
19Current breeding cycle 24 hours
New breeding cycle 6 hours
Re-scaling
24hrs
Up to 16-d
6hrs
Up to 16-d
Next T00Z
T00Z 10m
T00Z 40m
Re-scaling
Independent vectors
24hrs
Up to 16-d
Re-scaling
T06Z 10m
T06Z 40m
Up to 16-d
Re-scaling
24hrs
Up to 16-d
Re-scaling
T12Z 10m
T12Z 40m
Re-scaling
Up to 16-d
24hrs
Up to 16-d
Re-scaling
T18Z 10m
T18Z 40m
Re-scaling
Up to 16-d
20NCEPRecent ImplementationChanges - 3
- Relocation of perturbed tropical storm
- This change is intended to reduce track forecast
error and uncertainty for short lead-time (1-3
days) - Results
- Reducing mean track errors by 10 for 12-48 hours
- Reducing the ensemble track spread, that was too
large, for short lead-time - Improving track forecast skill
21GFS TS relocation
Ensemble TS relocation
6hrs fcst
Fcst/guess
3hrs
9hrs
6hrs
P
N
C
Use GFS Track information
Use ens. Track information
Use ens. Track information
Relocate TS to Observed position
Use GFS Track information
To separate into env. Flow (EF) And storm
perturbation (SP)
GDAS (SANL)
Ens. Rescaling For SP (pn)
Ens. Rescaling For EF (pn)
Combined
FCST
FCST
22Hurricane Track Plots (case 2)
Ivan (09/14)
Without relocation
With relocation
23Track error and spread2004 Atlantic Basin
(8/23-10/1)
From Timothy Marchok (GFDL)
Reduced mean track error and spread
24Hurricane track errors2 basins (Atlantic and
e-Pacific)
Percentage improvement to operational ensemble
Track errors (miles)
Period 20040824-20040930 (53-103 cases)
25North American Ensemble Forecast Systemsby
Zoltan Toth
- PARTICIPANTS
- PROJECT DESCRIPTION
- TIMELINE
- IMPLEMENTATION SCHEDULE
- CONCEPT OF OPERATIONS
- NAEFS THORPEX
- BASIC PRODUCTS
- END PRODUCTS
- DETAILS RESOURCE ISSUES
- FUTURE EXPANSION
- NEW NWP PARADIGM
- Visit http//wwwt.emc.ncep.noaa.gov/gmb/ens/NAEFS
.htm
26CONCEPT OF OPERATIONS
- Exchange 50 selected variables
- Use GRIB2 to reduce volume of data
- Generate basic products using same
algorithms/codes - Reduce systematic error
- Bias estimation
- Combine two ensembles
- Determine weights
- Express forecast in terms of climatological
anomalies - Prepare compare forecast with reanalysis
climate distribution - Generate center-specific end products
- Evaluate provide feedback for improvements
- Verification using same algorithms
- User feedback
- 2. MSC-NCEP basic production suite
- Same algorithms/codes used at both centers
- Duplicate procedures provide full backup in case
of problems at either end - If one component of ensemble missing, products
based on rest of ensemble - Basis for different sets of center-specific end
products
27NAEFS THORPEX
- Expands international collaboration
- Mexico joined in November 2004
- UK Met Office to join in 2006
- Provides framework for transitioning research
into operations - Prototype for ensemble component of THORPEX
legacy forecst system Global Interactive
Forecast System (GIFS)
RESEARCH
THORPEX Interactive Grand Global Ensemble (TIGGE)
THORPEX
RESEARCH
Articulates operational needs
Transfers New methods
North American Ensemble Forecast System (NAEFS)
OPERATIONAL
LEGACY (GIFS)
OPERATIONS
28Basic ProductsPost-Processing
- Bias corrected forecasts
- Consider 35 variables in the first phase
- Statistical weights
- Consider 35 variables in the first phase
- Anomaly forecasts
- Consider 19 variables in the first phase
- GRIB2
- NAWIPS grids and graphics
- NDGD grids
29Reviewing the Ensemble Prediction System By G.
Pellerin (CMC)
- In the currently operational EPS an ensemble
Kalman filter provides the initial conditions for
16 global 10-day forecasts at resolution of 1.2
degrees with two different dynamical models. - A new configuration in which the lead time is
extended to 16 days is being tested since the 2nd
of September. -
- We are planning to increase the number of
members to 20.
30Plan for the presentation
- description of the different components of the
EPS - the analyses with the Ensemble Kalman Filter,
- the 16-day medium-range forecasts using 2
models, - the essence of the changes tested with the
parallel run, - the impact of a new surface algorithm ISBA (
Interaction Soil Biosphere Atmosphere), - verifications of the near surface temperature,
- EPS exchanges with NCEP,
- future changes.
31ENSEMBLE SET-UP
observations
and
EnKF data assimilation
first guess fields
perturbed observations
6 hour integrations with the GEM model
perturbed analyses
perturbed fields for roughness length sea surface
temperature albedo
Selection of 16 analyses
Z0, SST, Alb
medium-range integration with the SEF model
medium-range forecasts
medium-range integration with the GEM model
32Changes to the analysis component
- addition of AMSU/A radiance data from AQUA,
- of MODIS derived winds from AQUA and TERRA, and
of dew-point spread at the surface. - changes in the assimilation cycles include the
use of a digital filter for the model, the
application of model error after the production
of the analyses, the breakup into 4 sub-ensembles
of 24 members (instead of 2 times 48 members), -
- preparation of the code for time-interpolation
by the ensemble Kalman filter.
33Changes to the forecast component
- motivated by the extension to day 16, required
simplified maintenance of model librairies and
required coherence of derived variables, - sharing of same more modern physical
parameterizations in both models, -
- application of a digital filter for all members,
-
- introduction of the ISBA surface interaction
algorithm.
34The operational set of perturbated model
configurations
Combination of modules for different model
perturbations
SEF (T149) Convection/Radiation GWD GWD
Orography Number Time level
version of levels 1 Kuo/
Garand Strong High altitude 0.3
23 3 2 Manabe/ Sasamori Strong Low altitude
0.3 41 3 3 Kuo/ Garand Weak Low
altitude Mean 23 3 4 Manabe/
Sasamori Weak High altitude Mean 41 3 5 Manabe/
Sasamori Strong Low altitude Mean 23 2 6 Kuo/
Garand Strong High altitude Mean 41 2 7 Manabe/
Sasamori Weak High altitude 0.3 23 2 8 Kuo/
Garand Weak Low altitude 0.3 41 2 control Kuo/
Garand Mean Low altitude 0.15 41 3 GEM
(1.20) Deep Shallow Soil
Sponge Number Coriolis convection
convection moisture of
levels 9 Kuosym new Less 20
global 28 Implicit 10 RAS old Les
s 20 equatorial 28
Implicit 11 RAS old Less 20
global 28 Implicit 12 Kuosym old M
ore 20 global 28
Implicit 13 Kuosym new More 20
global 28 Implicit 14 Kuosym new
Less 20 global 28
Implicit 15 Kuosym old Less 20
global 28 Implicit 16 OldKuo n
ew More 20 global 28
Implicit
35Review of SEF models
-
- removal of envelope orographies,
- use of a hybrid vertical coordinate (27 levels),
- introduction of a non-orographic GWD
parametrization, - replacement of Manabe with RAS convection
scheme, - use of a single condensation scheme (consun),
- use of the same radiation scheme (newrad) as in
GEM, - introduction of a new surface interaction scheme
(ISBA), - adjustment of the coefficients for horizontal
diffusion.
36Review of GEM models
- use of the same climatology as in EnKF model,
- introduction of a non-orographic GWD,
- introduction of new surface interaction scheme
(ISBA), - introduction of digital filter finalization.
37 The parallel set of perturbed model
configurations
SEF GWD Convection Schemes
Surface Number Time level
(T149) taufac deep
shallow scheme of levels Control
8.0e-6 Kuo conres
Fcrest 27 3
1 1.2e-5 Kuo conres
ISBA 27 3 2
1.2e-5 Ras turwet
Fcrest 27 3 3
4.0e-6 Kuo conres
Fcrest 27 3 4
4.0e-6 Ras
turwet ISBA 27 3 5
1.2e-5 Ras turwet
Fcrest 27 2 6
1.2e-5 Kuo conres ISBA
27 2 7 4.0e-6
Ras turwet ISBA 27 2
8 4.0e-6 Kuo
conres Fcrest 27 2 GEM
GWD Convection Schemes Surface
Number Time level (1.2) taufac
deep shallow scheme
of levels 9 8.0e-6
Kuosym ktrsnt Fcrest 28
2 10 8.0e-6
Ras conres ISBA 28
2 11 8.0e-6 Ras conres
Fcrest 28 2 12
8.0e-6 Kuosym ktrsnt ISBA 28
2 13 8.0e-6
Kuostd ktrsnt Fcrest
28 2 14 8.0e-6
Kuostd ktrsnt ISBA 28
2 15 8.0-e6
Kuosym conres ISBA 28
2 16 8.0e-6 Kuo conres
Fcrest 28 2
38New Hydrological Budget in ISBA
New Hydrological Budget in ISBA
Etr
RAIN
SNOW
Er
veg (1-psnv) Pr
psn Pr
LIQ. WAT. RETAINED ON THE CANOPY (Wr)
(1-veg)(1-psng) Pr
PS
ES
psn Rveg
(1-psn) Rveg
freezs
melts
SNOW (WS)
Eg
LIQ. WAT. IN SNOW (WL)
Rsurf
SOIL LIQUID WATER (w2)
Rsnow
meltg
freezg
FROZEN WATER IN SOIL (wF)
Drain
39Analysis of surface fields
observations
4D Var data assimilation
trial fields
6 hour integration the GEM model (fcrest)
analysis
6 hour integration GEM model (ISBA)
sfc trials (ISBA)
dynamic fields
Pseudo analysis
Surface fields ISBA scheme
new surface fields
40Pseudo-analysis of moisture
41MOGREPS The new Met Office short-range EPSby
Ken Mylne
- Ensemble designed for short-range
- Regional ensemble over N. Atlantic and Europe
(NAE) - Nested within global ensemble
- ETKF perturbations
- Stochastic physics
- T72 global, T36 regional
- Aim to assess uncertainty in short-range, eg.
- Rapid cyclogenesis
- Local details (wind etc)
- Precipitation
- Fog and cloud
NAE
MOGREPS is on Operational Trial for 1 year from
September 2005
42ETKF Initial Condition Perturbations
- ETKF Simplified version of Ensemble Kalman
Filter - ETKF similar to Error Breeding
- Perturbations are linear combination of forecast
perturbations from previous cycle, formed by
matrix transformation - Transforms calculated using same set of
observations as used in 4D-Var (including all
satellite obs) within /- 3 hours of data time
43ETKF Initial Condition Perturbations
- ETKF Simplified version of Ensemble Kalman
Filter - Cannot update mean state covariance
information only - Perturbations are added to 4D/3D Var analysis
44Stochastic physics
. the quest to increase spread!
Buizza et al., MWR, 2004
All three systems are under-dispersive!!
45Stochastic physics for the UM
- MOGREPS employs three schemes to address
different sources of model error - Stochastic Convective Vorticity (SCV)
- Unresolved impact of organised convection (MCSs)
- Not used in the higher resolution regional
ensemble - Random Parameters (RP)
- Structural error due to approximations in
parameterisation - Stochastic Kinetic Energy Backscatter (SKEB)
- Excess dissipation of energy at small scales
- SKEB not yet implemented
- Impact is propagated to next cycle through the
ETKF
46Stochastic schemes for the UM
The Random Parameters (Arribas, 2004) All
parameterizations include a number of
empirical-adjustable parameters and thresholds
(with somewhat arbitrary values!) These
parameters are treated as stochastic variables,
and, each 3-h, their values are calculated using
a first-order auto regression model
Ptµr(Pt-1- µ)e with r 0.95 Same value
at all grid points (i.e. spatial corr. 1)
47SKEB
Stochastic Kinetic Energy Backscatter (Arribas
and Shutts, 2005)
Aim To backscatter (stochastically) into the
forecast model some of the
energy excessively dissipated by it at
scales near the truncation limit. (similar to
ECMWFs CASBS by Shutts) A total dissipation of
0.75 Wm-2 has been estimated from the
Semi-lagrangian and Horizontal diffusion schemes.
Each member of the ensemble is perturbed by a
different realization of this backscatter forcing
48SKEB
Backscatter forcing
a.- Tunable amount of energy feedback KE.-
Kinetic Energy R.- Random field D.- Dissipation
rate ??.- Time-step
3D random pattern in which horizontal, vertical
and temporal correlations can be imposed to
reproduce CRM statistics
49SKEB
Preliminary results Positive increase in spread
(comparable to that seen at ECMWF)
Increase in spread respect to an IC-only
ensemble 500 hPa geopotential height
SKEB
RPSCV
50SKEB
Preliminary results Better representation of
forecast spectra
K-3
K-5/3
51MOGREPS Operational System diagram
New global analysis
Global ensemble forecast using stochastic physics
Perturbations mixed and scaled by ETKF
New NAE analysis
0Z
12Z
18Z
52Issues for discussion
- Relative benefit of different configurations
(ensemble size and resolution) - Probabilistic forecast (products development)
- Seamless forecast (day 1-7, week two)
- Forecast consistency
- Probabilistic evaluation
- Benefit to user
- Feasibility of developing sample overall measure
- Requirement for re-forecast