The Grey Zone Project: First Case: CONSTRAIN: A cold air outbreak A WGNE initiative
Research Progress Report 2007 - 2008...
Transcript of Research Progress Report 2007 - 2008...
Slide 1
Slide 1
Research Progress Report
2007 - 2008
ECMWF
Slide 2
Slide 2
Impact of forecast system upgrades2008
Nov 6th 2007 IFS Cycle 32r3 (partly discussed last year)
Jun 3rd 2008 IFS Cycle 33r1
Sept 30th 2008 IFS Cycle 35r1
Slide 3
Slide 3
Slide 4
Slide 4
Nov 6th 2007 IFS cycle 32r3
• Revision of the convection scheme, new formulation of convective
entrainment and relaxation time scale
• Reduction of vertical diffusion in inversions and free atmosphere
• New soil hydrology/runoff (HTESSEL)
• New radio-sonde temperature and humidity bias correction
• Increased amount of radio occultation data from COSMIC
• Assimilation of microwave AMSR-E, TMI and SSMIS window
channels
• Assimilation of ozone SBUV from NOAA-17 and NOAA-18.
• Reduce of initial perturbation amplitude for EPS by 30%, use new
moist physics package in computation of targeted tropical cyclone
singular vectors.
Slide 5
Slide 5
Normalized Anom Corr diffs
Slide 6
Slide 6
Tropical Forecast Biases Precipitation against GPCP for different cycles: from 15 year 5 months integrations for 1990-2005.
c 32r2 d 32r3
b 31r1a
Slide 7
Slide 7
More realistic model activity calculated from T799 deterministic forecasts against own Analysis
Cy32r3 significantly increases the level of activity in midlatitudes and Tropics
Slide 8
Slide 8
observation(for high pressure days over central Europe (last four winters))
DJF2004/558 cases
DJF2005/660 cases
DJF2007/869 cases
DJF2006/752 cases
ED
MF
PBL
M-O
diffu
sion
Slide 9
Slide 9
H-TESSEL The revised hydrology includes spatial variability related to
topography (runoff) and soil texture (drainage)
Slide 10
Slide 10
Improved soil moisture and evaporation
SEBEX (Savannah, Sandy soil)
BERMS (Boreal Forest)
HTESSEL improves soil moisture and evaporation with respect to TESSELin dry climates and leads to a better represented soil moisture inter-annual variability in continental climate
HTESSEL improves soil moisture and evaporation with respect to TESSELin dry climates and leads to a better represented soil moisture inter-annual variability in continental climate
Slide 11
Slide 11
Feb 3rd 2008 IFS cycle 33r1
• Improved moist physics in tangent linear/adjoint of 4D-Var.
• Physics: Retuned entrainment in convection scheme. Bugfix to
scaling of freezing term in convection scheme. Additional shear
term in diffusion coefficient of vertical diffusion. Increased turbulent
orographic form drag. Fix for soil temperature analysis in areas with
100% snow cover.
• Modified post-processing of 2m T and q.
• Active assimilation of AMSR-E and TMI rainy radiances.
• Use of 4 wind solutions for QuikSCAT.
• Extended coverage and increased resolution of limited area wave
model.
• Improved shallow water physics and modified advection for ocean
wave model.
Slide 12
Slide 12
Slide 13
Slide 13
enhanced orographic form drag
zonal drag[m2/s2]
RMS wind 850hPa[m/s]
33r1 e-suiteFeb-May 2008
Slide 14
Slide 14
Sept 30th 2008 IFS cycle 35r1 (aka 33r2)
OSTIA sea surface temperature and sea ice analysis
Conserving interpolation scheme for trajectory
New VARBC bias predictors to allow the correction of IR shortwave channels affected by solar effects
Cleaner cold start of AMSUA channel 14
New physics for melting of falling snow
Increased albedo of permanent snow cover
Cool skin/warm layer SST parametrization
Revised linear physics
Add convective contribution to wind gusts in post-processing
Monitoring of MERIS data
Slide 15
Slide 15
Interesting improvement of the model climate due to increase of Greenland and Antarctica albedo (from 0.75 to 0.8) (CY33R2)
CY33R1
Rev Snow
ERA-I
ERA-I
Long integrations (13-months) evaluated against several datasets indicate a consistent improvement (RMSE reduction)
Slide 16
Slide 16
Equitable threat score for European precipitation against SYNOP data
Curves show 12 month running mean of seasonal values
Calendar Years
1 day per 7 years
Slide 17
Slide 17
Slide 18
Slide 18
Large-scale eg MJO
Slide 19
Slide 19
Tropical Cyclone Intensity Error(mean of 365 days ending at 15 August)
-5
0
5
10
15
20
25
30
0 12 24 36 48 60 72 84 96 108 120
forecast step (hours)
core
pre
ssu
re (
hP
a)
2005
2006
2007
2008
ECMWF
Slide 20
Slide 20
Simulated Meteosat imagery
T799 36h forecast from 20080525
(Bechtold 2008)
Slide 21
Slide 21
Numerical Aspects Continuing optimization of the IFS (about 8% gained),
prepIFS, and various technical tools
In addition a more ambitious project to examine scalability of the IFS on ca 25,000 PEs has started
Continuing tests of the ALADIN non-hydrostatic dynamical core, in cooperation with Meteo-France More stable version developed, and alternative systems under analysis
Cost is 25 to 70% higher than hydrostatic version, more efficient version under study
Results and scores essentially identical for H and NH in the troposphere
In the stratosphere, NH results still inferior, due to lack of finite elements in NH (under development)
NH system under test in a great diversity of configurations (acoustic waves, mountain flows, convective flows, planetary waves)
Slide 22
Slide 22
The new “5-prognostic” cloud scheme
WATER VAPOUR
CLOUDLiquid/Ice
PRECIP Rain/Snow
Evaporation
Autoconversion
Evaporatio
n
Condensation
CLOUD FRACTION CLOUD
FRACTION
Current Cloud Scheme New Cloud Scheme
Slide 23
Slide 23
Model Ice Water Path (IWP) (1 year climate)
New 5 prognostic cloud microphysicsIce vs. Snow
CloudSat 1 year climatology
Current scheme with diagnostic snow
New scheme with prognostic snow
Observed Ice Water Path (IWP)
g m-2
Slide 24
Slide 24
Improvements in linear physics
Longwave radiation: TL/AD code has been developed and will replace soon the neural network approach
Allows more flexibility and variable trace gases
Several improvements in vertical diffusion
Mixed layer depth, fluctuations of diffusion coefficients, adapt to all changes in the NL model
Several improvements in convection
Much closer to NL code, important for rain assimilation
TL/AD of surface processes in development
Lake model (FLAKE) introduced in collaboration with Univ. Lisbonne and under testing, significant impact on surface evaporation expected
H-TESSEL run-off integrated to produce river discharges (TRIP model). Verifications against GRDC data allow to track problems.This module may be re-used by EC-Earth to couple river run-off and ocean salinity.
Representation of wind gusts under convection improved (problem signalled by DWD in May 2008)
Excessive snowfall in localized events during winter 2007/2008: problem identified (depth of the melting layer) and cured in cycle 35r1
As part of the work on assimilation of ground-based radar rainfall, a climatology of ducts and super-refraction has been produced , see www.ecmwf.int/products/forecasts/d/inspect/catalog/research/physics/ducting
Other on-going work on physics
Slide 26
Slide 26
Ocean waves
Excellent verifications of the maximum wave heights
Improvements in use of Quikscat (4 ambiguous solutions instead of 2, and new data stream from JPL with better quality control)
Monitoring of Jason-2 observations shows excellent quality
Offline experiments show positive impact of currents analyzed by Nansen Centre and Mercator-Ocean on wave heights simulations
Fully coupled use of currents, waves, and winds in IFS shows compensation by small changes in the wind fields (to keep comparable momentum fluxes) This influences the atmospheric forecasts and will require retuning of
some aspects of the physics
Slide 27
Slide 27
Limited area model (hindcast at 10km resolution): Hs difference (current – reference)
�41001
�41002�41004
�41008
�41009�41010
�41012
�41013
�41025
�41036
�41043
�41046
�41047
�41048
�42001
�42002
�42003
�42019
�42020
�42035
�42036
�42038
�42039�42040
�42041
�42054
�42055
�42056
�42057 �42058 �42059
�42099
�44004�44009
�44014
20°N 20°N
30°N30°N
40°N 40°N
90°W
90°W 80°W
80°W 70°W
70°WAnalysed wave height difference: currents (f05j) - no currents (ezy4)
ECMWF Analysis VT:Saturday 3 November 2007 00UTC Surface: **Significant wave height
-1.05
-0.9
-0.75
-0.6
-0.45
-0.3
-0.15
-0.010.01
0.16
0.31
0.46
0.61
0.76
0.91
1.05
snapshot
Slide 28
Slide 28
Average difference in surface wind
Current in wind direction:lower stress,but stronger surface wind
Slide 29
Slide 29
Data Assimilation The Ensemble Data Assimilation system is almost ready
for operational implementation
Initial configuration: 10 members+1 control
Incremental 4D-Var with T399 outer loops, T159 inner loops
Perturbed observations, spectral stochastic backscatter – benefits also from more active model since 32r3
Only short forecasts to provide a background for next cycle.
Spread is still a bit small, but impact on the EPS is positive, especially in the tropics. We will also learn from this initial operational configuration
Work will continue to base the deterministic assimilation on flow dependent background error variances computed from the EnDA. Benefits for severe weather forecasts are expected when this is achieved.
Slide 30
Slide 30
Day-to-day horizontal layer background error estimates for U-wind have very different structure and amplitude, calculated from operational randomization method and fromthe 2.0*(standard deviation of the 10 ensemble DA members). The ensemble based version is much more flow-dependent.
Operational randomization method 2.0*Stdev of 10 T255 outer loop+T95/T159 inner ensemble members.
Both ~850hPa
Slide 31
Slide 31
Hurricane Emily 19-20 July 2005
Ensemble Data Assimilation spread for zonal wind at 850hPa
Max. stdev of EnDA spread 19m/s Max. stdev of EnDA spread 30m/s
EPS probability at 00UTC 19 July
Slide 32
Slide 32
2
2
/ 2
/ 2H x
pk x k
if |x| <= k,
if |x| > k,
Introduction of the Huber norm
Gaussian
Huber
Gaussian + flat
Slide 33
Slide 33
The Huber-norm and robust estimation
18 months of conventional data
Feb 2006 – Sep 2007
Normalised fit of PDF to data
Best Gaussian fit
Best Huber norm fit
Slide 34
Slide 34
Comparing optimal observation weightsHuber-norm (red) vs. Gaussian+flat (blue)
More weight in the middle of the distribution
More weight on the edges of the distribution
More influence of data with large departures
Weights: 0 – 25%
25%
Slide 35
Slide 35
Changes in the IFS
Huber norm parameters for SYNOP, METAR, DRIBU: surface pressure, 10m wind
TEMP, AIREP: temperature, wind
PILOT: wind
Relaxation of the fg-check Relaxed first guess checks when Huber VarQC is done
Relaxation out to ~ 20 Sigma
Retuning of the observation error Smaller observation errors for Huber VarQC
Slide 36
Slide 36
Other work in progress on data assimilation Removing “balance constraints” (and perhaps gravity
wave penalty as well) seems to improve results – on-going tests
Technical developments for weak-constraint 4D-Var based on a 4D initial state (good scalability)
Scientific aspects of weak-constraint 4D-Var (e.g. need to represent time-correlated model errors) are studied with a simple 2-level QG model
Improvements of interpolations between outer loop and inner loop resolutions (shape-preserving interpolators)
Improvements in moisture assimilation to allow super-saturation in analysis and liquid and ice water in control variable
Preparation for operational Forecast Sensitivity to Observations suite
Slide 37
Slide 37
Satellite Data Progress in assimilation of rain-affected radiances Large improvements of the fit of TL physics to NL
physics
Improvements in description of scattering aspects in TL radiative transfer (included in RTTOV)
New instruments have been included (AMSR-E, TMI)
Periodic bias of 3K in TMI discovered and explained by 80K variations of temperature of the mirror. A bias correction has been proposed.
Direct 4D-Var is almost ready. Comparison between model and obs is done at model grid points.
Unification of clear and cloudy radiance assimilation in progress: single data flow and comparison at model grid points has been adopted, thinning now will be more consistent
Slide 38
Slide 38
Impact of 1D+4D-Var vs 4D-VarMean TCWV
analysis difference in %
1D+4D-Varminus
NORAIN
4D-Varminus
NORAIN
(08-09/2007)
Slide 39
Slide 39
33R1 without microwave imagers
33R1 control
All-sky 4D-Var without microwave imagers
All-sky 4D-Var
Relative Humidity RMS forecast errors
10th Aug to 10th Sept 2007: 24 to 32 samples verified against own analyses.
Slide 40
Slide 40
Higher resolution (horizontal and vertical):
- T1279 (EPS at T639) planned for later next year
- ~140-150 levels planned for 2010
One week of summer and one week of winter