Jean-Noël Thépaut ECMWF (and many colleagues from the Centre)

ECMWF19th NA-EUROPE data exchange meeting, 3-5 May 2006 Slide 1

ECMWF Status Report

Operational changes since 18th North America / Europe Data

Exchange Meetingand

short term plans

Jean-Noël Thépaut

ECMWF

(and many colleagues from the Centre)


OUTLINE

Cycle 29R2 (28 June 2005)

Cycle 30R1 (1 February 2006). The higher-resolution forecasting system T799/L91, T255 4D-Var

Other changes

Short (and long) term plans


Model Cycle 29R2 (28 June 2005)

Initial 1D-Var+4D-Var assimilation of SSM/I rain affected radiances

Changes to AIRS assimilation

Introduction of Meteosat-8 (MSG) winds

New Jb stats (from recent ensemble of data assimilation)

Modifications to humidity analysis (less increments in areas with high CAPE, reduced spinup)

Modifications to convection scheme (bugfix for negative mass flux, implicit momentum+tracer transport)

Revision of the initial perturbations for the EPS

Introduction of SMHI Baltic sea ice


1D-Var + 4D-Var assimilation of SSM/I rain affected radiancesThe 1D-Var observation operator developed for this project is

the most complex ever, and the good (Gaussian) distribution of (obs-background) is a remarkable result

The change resulted in a small global moistening and a reduction of the tropical precipitation spin-down

Comparisons with independent estimates of the TCWV (Jason) are improved

Short-term forecasts of humidity improve in the tropics; The effect is still limited otherwise (w not part of the control variable)


1D-VarObservation operator:Lin. Large-scale condensationLin. ConvectionRTTOV-SCATT

Post-screening

SSM/I TB’s

Scan-biascorrection

Interpolationto model grid

Pre-screening

TCWV-observation4D-Var

Air-mass bias correction

Implementation


Assimilation of rain-affected microwave radiancesand improvement of humidity analysis

e-suite ops

Global 1.74 1.90N. Hemisphere 1.63 1.71Tropics 2.12 2.43S. Hemisphere 1.53 1.62N. Atlantic 1.63 1.69N. Pacific 1.57 1.69

stdev (kg/m2)

Comparison of cycle 29r2e-suite and operations

with independent TCWV retrievals from Jasonmicrowave radiometer

(may 2005)


used qTEMP-q Caribexp:0001 2005082300-2005083012(12)

nobsexp

1474

1871

2982

3680

2815

1683

519

417

28

1

0

exp - ref

-4

-2

+2

+1

-4

+0

+0

+0

+0

+0

+0

0 0.001 0.002 0.003 0.004

STD.DEV

1000

850

700

500

400

300

250

200

150

100

70

Pre

ssur

e (h

Pa)

-0.0006 0 0 0 0

BIAS

1000

850

700

500

400

300

250

200

150

100

70

background departure o-b(ref)background departure o-banalysis departure o-a(ref)

analysis departure o-a

used UDrop-windspeed Caribexp:0001 2005082300-2005083012(12)

nobsexp

304 267 342 345 331 301 207 230 111 0 0 0 0 0 0 0

exp - ref

-10 -16 -3 +0 +0 +0 +0 +0 +0 +0 +0 +0 +0 +0 +0 +0

0 2 4 6

STD.DEV

1000 850 700 500 400 300 250 200 150 100 70 50 30 20 10 5

Pre

ssur

e (h

Pa)

-4 -3 -2 -1 0 1 2 3 4

BIAS OF SPEED

1000 850 700 500 400 300 250 200 150 100 70 50 30 20 10 5



used TDrop-T Caribexp:0001 2005082300-2005083012(12)

nobsexp

421 427 615 647 531 303 181 148 103

0 0 0 0 0 0 0

exp - ref

+0 +1 -8 +0 +0 +0 +0 +0 +0 +0 +0 +0 +0 +0 +0 +0

0 0.5 1 1.5 2 2.5 3

STD.DEV

1000 850 700 500 400 300 250 200 150 100 70 50 30 20 10 5

Pre

ssur

e (h

Pa)

-2 -1.5 -1 -0.5 0 0.5 1 1.5 2

BIAS

1000 850 700 500 400 300 250 200 150 100 70 50 30 20 10 5



CY29R2: 0001CY29R2 no TCWV: exp

Area:

TC Katrina Dropsonde Departure Statistics: 20050823-20050830

Radiosonde:Specific humidity

Dropsonde:Windspeed

Dropsonde:Temperature

1004

1008

1008

1008

1012

1012

1016

5

10°N10°N

20°N 20°N

30°N30°N

90°W

90°W 80°W

80°W 70°W

70°W

Friday 26 August 2005 00UTC ECMWF Forecast t+48 VT: Sunday 28 August 2005 00UTC Surface: **large scale precip/Surf: mean sea level pressure/850hPa vorticity

0.1

0.5

1

2

3

5

10

20

50

100



retune existing bias model coefficients

new surface emissivity model

intelligent thinning of AIRS data

new observation errors (long-wave channels)

VIS / NIR cloud information

VARBC technical changes



Impact of changes :

•Significant increase in tropospheric AIRS data use (mostly due to thinning and emissivity model)

•Small forecast improvement

•No change to stratospheric oscillations

Stratospheric channels usage unchangedTropospheric /

surface sensing channel usage increased

Channel wavelength


Bias: METEOSAT-8, ekip_FGQI: 2004120200-2004120418WVcloud1: N = 314252.

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

WVcloud2: N = 343138.

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

IR3: N = 376360.

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

VIS2: N = 92955.0

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

WVMV1: N = 83640.0

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

WVMV2: N = 93692.0

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

Bias: METEOSAT-8, ekip_FGQI: 2004120200-2004120418WVcloud1: N = 314252.

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

WVcloud2: N = 343138.

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

IR3: N = 376360.

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

VIS2: N = 92955.0

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

WVMV1: N = 83640.0

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

WVMV2: N = 93692.0

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

Bias: METEOSAT-8, ekip_FGQI_used: 2004120200-2004120418WVcloud1: N = 43360.0

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

WVcloud2: N = 36934.0

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

IR3: N = 26462.0

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

VIS2: N = 18467.0

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

Bias: METEOSAT-8, ekip_FGQI_used: 2004120200-2004120418WVcloud1: N = 43360.0

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

WVcloud2: N = 36934.0

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

IR3: N = 26462.0

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

VIS2: N = 18467.0

-60 -40 -20 0 20 40 60Lat [deg]

1000

800

600

400

200

p [h

Pa]

-7.0 -6.0 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0

IR10.8 µm WV6.2 µm WV7.3 µm

QI>60

“USED” WINDS Blacklist forms part of selection:QI>85IR: no winds between 700-460hPaWVcloudy 6.2: no winds below 400hPaWVcloudy 7.3: no winds below 600hPaWVcloudy and IR: no winds below 250hPa level in the Tropics (25S-25N)VIS: no winds above 700 hPa2-4 Dec 2005

An assimilation experiment was carried out for 89 days. Forecast scores for up to forecast day 10 showed a neutral overall impact: NH: slightly positive to neutral SH: slightly negative to neutral Tropics: neutral

Introduction of MSG (Meteosat-8) AMVs


29R2: Operational change in June 2005

Scores positive in Europe (99% statistical

significance at day 3 and 98% at day 7)

Scores positive in Northern Hemisphere (99%

statistical significance at day 3 5 and 7)

Scores neutral in Southern Hemisphere

Day

Anomaly correlation of 500hPa forecasts for Europe

Mean from 1 Dec 2004 to 28 June 2005

Cycle 29r2Cycle 29r1


The 30r1 suite, T799/L91 (1 Feb 2006)

T799 horizontal resolution (12min time step)4D-Var increments at T255 (30min time step)Vertical resolution increased to 91 levelsModel top raised to 0.01hPaGrid-point humidity and ozone in 4D-Var Changes to the wave model

- Resolution increased from 0.5° to 0.36°

- Use of Jason altimeter wave height data and ENVISAT ASAR spectra in the wave model assimilation. ERS-2 SAR spectra no longer assimilated.

Changes to EPS- Resolution increased from T255L40 to T399L62 (30min time step).

- Wave model resolution unchanged at 1° , but number of frequencies increased from 25 to 30, and number of directions changed from 12 to 24.


Three problems encountered

1) Humidity at the top of the model (in the mesosphere) increased slowly in data assimilation1) Solved by scaling down the humidity increments

2) Occasional instabilities of the TL (over steep orography)2) Addressed by more precise handling of the NL trajectory

3) A strong jet develops in the tropical mesosphere 3) Addressed by retuning Rayleigh friction

Parallel runs restarted with 30r1 Altogether (RD + OD) the whole period from 1st July to

change date + Summer 2004 ~300 cases


Statistical significance, t-test, 500hPa height

Green denotes T799L91 better (red) would denote T799L91 worse

Day 3 Day 5 Day 7

N HemACRMS

0.5%

0.1%

-

2%

-

-

S HemACRMS

0.1%

0.1%

0.1%

0.1%

0.2%

0.1%

EuropeACRMS

0.5%

0.1%

5%

0.1%

-

-

260 00UTC cases from 1 January 2005


Other statistics from the high resolution suite

More observations are accepted by the data assimilation in the E-suite (e.g. 20% more synop winds!)

The background and the analysis are in average closer to synop winds

The observation cost function at the end of the minimization is smaller in the E-suite despite of the increased number of observations

Forecasts from the E-suite verify better against radio-sonde winds, especially for Europe and the Tropics

Ocean wave forecasts from the E-suite are substantially better


Forecasts of Katrina for 12 UTC, Monday 29 August

OperationalT511 L60

72h forecast

36h forecast

OperationalT511 L60

TestT799 L91

TestT799 L91

+

++

+


Other changes over the last year

Use of warmest FOV of AIRS

NOAA-18

- HIRS operationally monitored

- AMSU-A and MHS operationally assimilated

Operational monitoring of MTSAT-1R AMVs

New bias correction coefficients implemented on February 8


NOAA-18 AMSU-A in operations

Radiances from the NOAA-18 AMSU-A instrument have been monitored / bias corrected and were successfully introduced to operations on the 8th September

NOAA-18 AMSUA channel 8 (O-B)

Passive monitoring OPS


NOAA-18 HIRS/MHS passive monitoring

Data from the NOAA-18 HIRS and MHS instruments have beenpassively monitored and bias corrections have been established. Impact of MHS in assimilation mode revealed neutral to slightly positive.

NOAA-18 HIRS channel 4 NOAA-18 MHS channel 5

Monitoring of the NOAA-18 MHS show that the radiances are less noisy than those from the AMSU-B (confirming improved instrument design for METOP)

NOAA-16 AMSUB channel 5 NOAA-18 MHS channel 5

Standard deviation of (O-B)


Active data

Z anomaly correlation scores:

OVERALL NEUTRAL impact. Only significantly positive result: NH 200hPa – day 3.

MTSAT: An assimilation experiment was carried out for 31 days (Aug 2005). Data selection same as GOES-9 but with QI > 85. Results below are using JMA’s forecast dependent QI.

1-31 Aug 2005

80

60

40

20

0

wind speed

0 20 40 60 80FG

0

20

40

60

80

SAT

OB

IR

(

1

953)

5

55

10

10

30

lat: -90 - -20

lon: 60 - -40

p : 0 - 400

SH

0 20 40 60 80

FG

OB

S I

R (

1953

)

80

60

40

20

0

wind speed

0 20 40 60 80FG

0

20

40

60

80

SAT

OB

IR

(

8

726)

5

5

10

10

307515

0

lat: -20 - 20

lon: 60 - -40

p : 0 - 400

Tropics

0 20 40 60 80

FGO

BS

IR

(87

26)

80

60

40

20

0

OB

S I

R (

4751

)wind speed

0 20 40 60 80FG

0

20

40

60

80SA

TO

B I

R(

475

1)

5

5

5

10

10

30

75

5

lat: 20 - 90

lon: 60 - -40

p : 0 - 400

0 20 40 60 80

FG

NH

WIND SPEED

0 24 48 72 96 120 144 168 192 216 240Step

-0.1

-0.05

0

0.05

0.1

0.15

Me

an

Diff

ere

nce

/ (1

00

% -

Me

an

Sco

re)

eqdd minus erbk Z 500 N.HEM ANCF90.0% two sided confidence intervalControl-experiment Z 500hPa

0.15

0.1

0.05

0

-0.05

-0.10 1 2 3 4 5 6 7 8 9 10

daysMea

n d

iffe

ren

ce /

(100

% -

mea

n s

core

)

0 24 48 72 96 120 144 168 192 216 240Step

-0.1

-0.05

0

0.05

0.1

0.15

Me

an

Diff

ere

nce

/ (1

00

% -

Me

an

Sco

re)

eqdd minus erbk Z 200 N.HEM ANCF90.0% two sided confidence intervalControl-experiment Z 200hPa

0.15

0.1

0.05

0

-0.05

-0.10 1 2 3 4 5 6 7 8 9 10

daysMea

n d

iffe

ren

ce /

(100

% -

mea

n s

core

)


8 February 2006 – New radiance bias correction

Applied statically, but derived from variational scheme to be implemented with cycle 30r2

pre

ssu

re (

hP

a)

Sonde-bg ContolSonde-bg New bias correctionSonde-an ControlSonde-an New bias correction

Temperature (K)

N Hem

RMS error of 300hPa tropical temperature forecasts

New bias correction

Control

Mean from 8-31 Jan 2006

Day


Data from twenty-eight sources used daily

Large increase in number of data used daily

Use of satellite data today


Short term plans

EPS extension to day 15- T255 perturbed forecasts from day 10 to day 15

- T399/255 control to day 10/15

- Also uniform T399 control to day 15 and T255 control to day 21

Variational radiance bias correction

Thinning of low-level AMDAR data

Improved treatment of ice sedimentation and autoconversion to snow in cloud scheme

Implicit treatment of convective momentum transport

To be used for T255L91 interim reanalysis for 1989 - ….To be used in version 3 of Seasonal Forecasting System?


Also for 2Q or 3Q 2006:

Introduction of turbulent orographic drag scheme

Use of albedo fields from MODIS

Use of high-resolution NCEP SST fields

Refinements to stratospheric analysis

Recalibrated radiosonde temperature bias corrections


And for 4Q 2006:4D-Var changes:

- 3rd inner loop- revised trajectory interpolation- revised data usage, including modified Var QC- new cloud and convection schemes in minimization

Upgrade fast radiative transfer to RTTOV-9

Change model short-wave radiation scheme to RRTM-SW

Unified medium-range/monthly EPS

Start of GEMS constituent reanalyses for 2003/4

And for the use of new satellite data:Monitoring and later assimilation of data from:

- AMSR-E, TMI and SSMIS (for clear-sky and then rainy conditions)

- CHAMP, COSMIC and GRAS

- FY-2C, Met-9

- METOP ATOVS + preparation for IASI

- Preparation for GOME-2 and ASCAT


GPS RO (CHAMP) assimilation trials 1D and 2D obs operators, June/July 2004

Polar stratospheric bias and oscillations are largely improved by GPS data

Operational assimilation of CHAMP data will start as soon as data are received in real-time (expected end 2005), then COSMIC and GRAS


Longer-term elements of four-year plan:

Ensemble component to data assimilation

Explore long-window weak-constraint 4D-Var

Exploit METOP, ADM, SMOS and other new satellite data

Continue refinement of physical parametrization

Assess need and plan if required for new dynamical core

Coastal zone model for ocean waves and storm surges

Prepare for next resolution upgrade


Longer-term elements of four-year plan:

New stochastic physics and high res moist SVs for EPS

Couple ocean and sea-ice throughout unified EPS

Develop OPA-based Seasonal Forecasting System

Establish GEMS system for constituent analysis

Continue interim reanalysis and prepare for ERA-65/75


The GEMS project has started

Funded by European Commission under Framework Programme 6 (one of the GMES IPs)

ECMWF and 25 partner institutes in Europe

Deliverables:

- An operational capability for global analyses and forecasts of greenhouse gases (CO2, CH4, N2O), reactive gases (O3, NO2,SO2, CO, HCHO) and aerosols (5 categories initially)

- A reanalysis of the recent period and validation against in situ measurements

- Use as much satellite and in situ data as possible

- Provide boundary conditions for regional air quality models and organize a European-size inter-comparison between regional air quality models


CO2 Comparison with flight data from Japan Air Lines

Flight data kindly provided by H. Matsueda, MRI/JMA


The interim reanalysis- will start soon with T255L91, 4D-Var and cycle 30r2 (?)

- will run from 1989 onwards, and be continued in close to real time

- will use mostly same pre-2002 data as ERA-40, but will include

reprocessed winds from EUMETSAT

reprocessed ERS altimeter data

GOME profile data from RAL

improved radiosonde bias corrections

European Regional Reanalysis (EURRA)

- potential for a European regional reanalysis project (EURRA) is being explored with NMSs and EEA

A new major reanalysis (ERA-65 or -75) in due course- funding has to be secured

ECMWF plans for reanalysis


Much improved 500hPa height forecasts

Day

Southern hemisphere

Anomaly correlation of 500hPa height, averaged for 12UTC forecasts from 1 January to 31 December 1989

ERA-new

ERA-40

Operations

Northern hemisphere


ECMWF Strategy for 2006-2015The principal goal of ECMWF in the coming ten years will be

to maintain the current, rapid rate of improvement of its global, medium-range weather forecasting products, with particular effort on early warnings of severe weather events

Complementary goals are:- To improve the quality and scope of monthly and seasonal-to-

interannual forecasts

- To enhance support to Member States national forecasting activities by providing suitable boundary conditions for limited-area models (and regional CTMs)

- To deliver real-time analyses and forecasts of atmospheric composition

- To carry out climate monitoring through regular reanalyses of the Earth-system

- To contribute towards the optimization of the Global Observing System


Thank you

Documents

Jean-Noël Thépaut ECMWF (and many colleagues from the Centre)