1 10/2003 Crown copyright Unified Model Developments 2003 Clive Wilson NWP Met Office.

1 10/2003 © Crown copyright

Unified Model Developments 2003

Clive Wilson

NWP

Met Office


Outline

New location Configurations & Users Global and UK Mesoscale Changes Regional European Model Met.no high resolution (3km) New supercomputer - 15 node NEC SX6 Plans


New Headquarters & Operations Centre

24 September 2003


21 August 2003


Specifications

13,000 square metres of office space £80m two IT halls, two supercomputers interactive innovation centre media centre training college library conference facilities restaurant sports and recreation area an observations trial site to develop

and test new systems

Exeter

Bracknell


Operations centre


Operational Versions & External Users

Global 38 levels, N216= 432x325 (~60km)

UK Mesoscale 38 levels, 0.11 deg (12km)

Stratospheric 42 levels (~1.3km @ 100-1hpa), N48=96x73 (~300km)

Crisis models (LAMS) Middle East, Former Yugoslavia,SW Asia

External CGAM UK, ICM Warsaw, Thai Met Service, New Zealand, met.no


Changes- Global G29 3 Dec 02 Merge main/preliminary forecast runs

1h 50 min data cut-off digital filter initialisation instead of incremental update

G30 11 Dec Tile surface scheme with single tile (MOSES II) Quikscat (SEAWINDS), BUFR Meteosat winds correction of time interpolation for pressure pre-

processing G31 11 March 03

shallow convection w<0 target 2 moisture diffusion w>0.2ms-1


Estimated Skill score change for extended "Preliminary" Global model. Period =200012-200111

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

0.04

0.05

T+24PMSL

NH

T+48PMSL

NH

T+72PMSL

NH

T+96PMSL

NH

T+120PMSLNH

T+24H500NH

T+48H500NH

T+72H500NH

T+24W250

NH

T+24W850Trop

T+48W850Trop

T+72W850Trop

T+24W250Trop

T+24PMSL

SH

T+48PMSL

SH

T+72PMSL

SH

T+96PMSL

SH

T+120PMSLSH

T+24H500SH

T+48H500SH

T+72H500SH

T+24W250

SH

Parameter

unweighted skill diff Wtd Skill diff

Reduced data cut-off - Estimated Global worsening

250hPa winds

estimated


Annual Skill scores


Digital filter initialisation-Adam Clayton

Perturabation Forecast model used to propagate analysis increment forwards/backwards

States digital filtered Initialised PF analysis RH increments restored Interpolate to model resolution


Spin upMoisture Spin up

Quickly adjusts to initialised state


Shallow convection (Alan Grant/ Sean Milton ) -removal of spurious subtropical cloud


Targeted moisture diffusion

Max velocity (Sept 6th)

0

2

4

6

8

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

time (hrs)

m/s Control

Test

Previously reduced 0.5*timestep to control grid point storms

w based diffusion of moisture

2 moisture diffusion locally where w>0.2ms-1

15% fewer grid points with high w


Changes- limited area– UK mesoscale M25 4 Dec 02

Halve weights for MOPS cloud assimilation retune covariances

M26 18 Mar 03

Reduced latent heat nudging Introduce AMSU target 2 moisture diffusion w>1.0 ms-1

European LAM - Dec 02 introduced forecast only 06/18UTC Aug 03 with data assimilation


Dec 02 data assimilation: retuned covariancesreduced weight to MOPS cloud


Reduced Latent Heat Nudging

control upgrade


Reduced Latent Heat Nudging

controlupgrade


AMSU impact

Without With


European Model

20km, 38 levels Introduced Dec 02 semi-operational (06/18UTC) Data assimilation from 12 Aug Lateral bc & initial data => met.no Plan : 4 runs /day, 50 levels Q2/2004

12km resolution Q4/2004 60/70 levels Q2/2005



Norway Co-operation

3km resolution European model

analysis & lateral bc met.no running semi-

operational from July 03– from 10km Hirlam

– improved land/sea mask

– improved orography


Norway 3km


High res orography influence on precipitation


High resolution - JCMMPeter Clark, Humphrey Lean, Richard Forbes


Timetable Installation

Sep 02Oct 02Nov 02Dec 02Jan 03Feb 03Mar 03Apr 03May 03Jun 03Jul 03Aug 03Sep 03Oct 03Nov 03Dec 03

Date Bracknell SX-6

1 node syste

m

Upgrade:4 nodes

Front end

Move to Exeter

T3E-A

Move to Exeter

02/09-06/10

T3E-B

Move to Exeter

17/06-22/07

ExeterSX-6(B)

15 node system

ExeterSX-6(A)

15 node system

Acceptance tests


Met Office Networks

FC S

wit

ch

Mirrored filesystemsUser

filesystems

FibreChannel

Fron

t En

d12

xIA6

4

Fron

t En

d12

xIA6

4

Gigabit Etherne

t

The BitsN

OD

E(8

CP

Us)

NO

DE

(8

CPU

s)N

OD

E(8

CP

Us)

NO

DE

(8

CPU

s)N

OD

E(8

CP

Us)

NO

DE

(8

CPU

s)N

OD

E(8

CP

Us)

NO

DE

(8

CPU

s)N

OD

E(8

CP

Us)

NO

DE

(8

CPU

s)

NO

DE

(8

CPU

s)N

OD

E(8

CP

Us)

NO

DE

(8

CPU

s)N

OD

E(8

CP

Us)

NO

DE

(8

CPU

s)

IXS(Interconne

ct)

IXS


Plans 2004- 2005 Port to SX6 - Operational March 2004 Add stratosphere to Global, Euro-,UK models

– 50 levels, 2Q/2004

Global 4D-var 4Q/2004 12km European model 4Q/2004 Global 40km, 60 L 2Q/2005 European 60/70 levels 2Q/2005 4km UK 4Q/2005


Incremental 4D-Var


The PF (Perturbation Forecast) Model PF model is linearization of the full model (non

hydrostatic) with some small terms omitted. Simplified linear physics. Predicts evolution of finite perturbations with respect to a basic state.

It is not TL (tangent linear) to any particular model. More flexibility and allows cost saving approximations.

Avoids problems with linearizing semi-Lagrangian interpolation and solver for elliptic equation

Adjoint of PF model coded explicitly.


Trial 4D-Var System Trial suites with single or multiple outerloops at

N144/N108 , tests at N48/N48 Time window T-3 to T+3, obs used at nearest

timestep and processed as for 3D-Var with high res, time interpolated fc for innovation

Low resolution trajectory interpolated from high resolution fc, available every timestep

PF adiabatic plus momentum mixing PF initialization


12UTC 31 Dec 200124hour fc -

analysisof mean sea

level pressure

Rick Rawlins, Adam Clayton, DingMin Li etal

3D-Var

4D-Var


Plans Complete development of variable resolution multiple-outerloop,

test more physics Global 4D-Var operational 2004 Complete development of limited area version Develop ability to assimilate cloud and precipitation European 12km Lam 4D-Var 2005 Physics

—new gravity wave drag

—new convection

—new cloud (prognostic cloud and condensate)

1 10/2003 Crown copyright Unified Model Developments 2003 Clive Wilson NWP Met Office.

Documents