unified model developments 2005 for ewglam/srnwp annual meeting 2005

© Crown copyright 2004

Unified Model Developments 2005for EWGLAM/SRNWP Annual Meeting 2005

3-5 October 2005, Ljubljana, Slovenia

Mike Bush

NWP


1. Current UM configurations and future plans

2. Global model

3. Global and regional ensembles

4. NAE (North-Atlantic European) model

5. CAMMS and Africa model

6. HRTM

7. U.K 4km model


Global60KmNAE

12Km

UK Mes12Km

Current NWP Model Configurations

CAMMs

38 levels

UK4km

Africa20km

Trial ensembleGlobal 90km & Regional 24km


Global40Km

NAE12Km

2006: NWP Model Configurations

CAMMs

70 levels

UK4km

Africa20km

Trial ensembleGlobal 90km & Regional 24km


Supercomputers for NWP and Climate

NEC SX6 Operational since April 2004Twin 15 node system 1.86 TFlop

April 2005 UpgradeAdditional 4 SX6 nodesPlus 16 node NEC SX84.08 TFlop

Will allow the running of ensembles and increased resolution models including those used in climate research



2. Global model




6. HRTM

7. U.K 4km model


Global model developments

G33 5th October 2004 Introduction of 4D-VAR.

G34 18th January 2005 HadGEM physics upgrade

G35 8th February 2005 Data Assimilation upgrade.

G36 14th June 2005 DA and satellite upgrade.

G37 17th August 2005 Implementation of a soil moisture nudging scheme.


Global model cycle G34 – 18th Jan 2005

Introduction of HadGEM physics3C Large Scale precipitation scheme (replacing 3B).8B Boundary Layer (replacing 8A). Increase in Saharan surface albedos.

Benefits Improved tropical performance

o Increased (Reduced) Precipitation over land (Ocean).o Improved tropical circulation & winds.

Reduced low cloud (Iraq); Improved 1.5m T. Improved surface T and circulation over Sahara.

Drawbacks Increase existing warm bias in extratropics


Systematic Errors in Tropical Precipitation


Impact of 30min CAPE closure


Global model cycle G37 – 17th Aug 2005

Implementation of a soil moisture nudging scheme

Replaces the weekly resetting to climatology.Errors in the 6-hour forecasts of screen temperature and

humidity are used to infer corrections to the soil moisture field.

The scheme derives these from the physics of the land-surface model.

Summer trials showed improvements to surface and boundary-layer temperature forecasts.

Winter results were close to neutral.


Planned Global Changes (2005-6)

1. 4D-Var upgrade • better incremental physics, less obs thinning (ATOVS)

2. Increased Resolution (phase 1)• ~ 40 km, ~50 level

3. Better use of Satellite Data• SSMI-S, Microwave Cloudy radiances, MSG winds (Meteosat 7 -> 8)

4. Convection and Boundary Layer tuning

5. Increased Resolution (phase 2) • -> 70 levels

2005Q3 2005Q4 2006Q1 2006Q2

1 2 3 4 5



2. Global model




6. HRTM

7. U.K 4km model


Introduction of Ensembles

1. GEPS1 14th June 2005 24 member Global Ensemble at N144 (~90km) resolution twice per

day (00Z, 12Z) to T+72 hours. ETKF perturbations, stochastic physics Initially there will be a one year trial in which forecasters can view

output on the internal web

2. EPS1 17th August 2005 24 member North Atlantic/European Ensemble at 24km resolution

twice per day (06Z, 18Z) to T+36 hours IC perturbations taken directly from global model Nested within global ensemble for LBCs

3. Plans to engage in multi-model ensembles with USA and Canada (THORPEX). Proof of concept by March 2006


MOGREPS Spaghetti chart


MOGREPS Probability map



2. Global model




6. HRTM

7. U.K 4km model


NAE model developments

E5 18th January 2005Change in domain

E6 22nd February 2005Horizontal resolution increased from 20km to 12km.

E7 14th June 2005New surface soil moisture analysis.

E8 17th August 2005Removal of truncation of vertical modes in VAR

from 21 to full 38.


Planned NAE Changes (2005-6)

1. Use of modified Global covariances

2. Introduction of HadGEM physics

3. Introduction of 4D-Var

4. More new satellite data• Full resolution AMSU-B, Assimilation of GPS data, MODIS winds etc.

5. Increased Resolution • 70 levels

2005Q3 2005Q4 2006Q1 2006Q2 1 2 3 4 5


NAE model cycle E9 – 12th Oct 2005

Use of modified Global covariances

Generated using the NMC method

Replaces the use of U.K Mes covariances.

Global covariances fit less closely to obs but the horizontal length scales are longer.

Large improvements to the performance of the model

Ideally would use NAE covariances, but forecast differences are dominated by the use of different lateral boundary conditions.


NAE PMSL bias and RMSE


NAE 10m wind speed bias and RMS Vector Error


NAE Screen level temperature bias and RMSE


NAE performance issues

The NAE model (12km resolution) covers the North Atlantic – it is unusual for a LAM to cover such a large sea area.

Rationale is that by modelling the North Atlantic at high resolution, forecasts for Europe will be better than if a smaller domain over Europe was chosen.

Choice of covariances of crucial importance to the model performance (must do a good job over both sea and land).

The Global model does a good job at synoptic scales and the NAE has struggled to do a better job than it at these scales (lateral boundary conditions don’t help either).

LAM’s are good at adding detail to Global model output.



2. Global model




6. HRTM

7. U.K 4km model


Current Operational CAMM configurations

Southern Asia CAMM

Southern Asia (17km)

Falklands (12km)


Bay of Bengal CAMM (Tsunami Aid effort)

Bay of Bengal CAMM

17km 00z and 12z T+48320x240 38 levels

Requested Tuesday 4th JanDelivered to Operations Thursday 6th JanOperational Monday 10th Jan


USA CAMM: Hurricane Rita

Global Model (60km) USA CAMM (17km)

70kt 10m wind, 962mb

Global model central pressure: 992mb USA CAMM central pressure: 972mb NHC advisory (estimate): 929mb


USA CAMM: Hurricane Rita

Global Model (60km) USA CAMM (17km)

70kt 10m wind, 962mb

USA CAMM has track slightly west of the Global model


African LAM

Met Office vision: increased focus on assisting developing countries / disaster mitigation

Introduced into the Operational Suite on 13th April 2005

Currently one forecast a day (00Z) to T+48

20km horizontal resolution

Contains HadGEM physics

Introduction of data assimilation (6 hour cycle) in Winter 2005/2006

Increased vertical resolution (2006)


African LAM


Africa LAM rainfall on 16th April 2005

TRMM =Tropical Rainfall Measurement Mission satellite



2. Global model




6. HRTM

7. U.K 4km model


HRTM Model

Centred Over the Chilbolton Radar (High Resolution observational data for evaluation and validation)

HRTM 4 km nested in operational UK-MES. Used for research and to provide lateral boundary conditions to the HRTM 1 km

Over two years in development

Basis for the now operational UK4 model

Continued development

High Resolution Trial Model


HRTM Scientific Options

Dynamics: Non hydrostatic, Semi-Lagrangian advection and Semi-Implicit time integration (New Dynamics).

Boundary Layer: 1st order non-local K scheme with explicit entrainment.

Horizontal diffusion: Del-4 operator in U, V, Q, and Theta.

Microphysics: Mixed phase scheme + 3D advection of precipitation products.

Radiation: Two stream scheme with 5 spectral bands for short wave and 5 for long wave


HRTM Scientific Options (Convection)

Deep convection: explicitly resolvedShallow convection: parametrised

Solution based on the operational mass flux scheme with CAPE closure

CAPE reduction timescale dependent on CAPE.Large values of CAPE: CAPE reduction timescale increases. Activity of convection scheme reduced and deep convection is explicitly resolved.Small CAPE: Minimum value of CAPE timescale fixed. Shallow convection processes are taken into account.

Delay in the onset of convection as the convection scheme is made less active.


HRTM Scientific Options (Convection)

Function chosen: linear with CAPE for large CAPE values and exponential for small CAPE values

min governs activity of convection scheme for small CAPE values

min/CAPEmin governs the asymptotic slope of the function and is related to the limit to the cloud base mass flux


HRTM Tests: Effect of lateral boundary updating frequency

60 minutes 30 minutes 15 minutes

Chosen frequency: 30 minutes



2. Global model




6. HRTM

7. U.K 4km model


UK4 Model

320 rows by 288 columns

0.036 degrees gridlength (4 km aprox.)

38 Levels (13 in boundary layer)

100 seconds timestepSame scientific settings

as HRTM


UK4 Model (Pre-operational tests)

21 Case studies:Coverage of all main weather regimes, although biased towards severe weather events.Nested in UK-MES (12 km)

Continuous trial1st of March to 1st of April 2005, at 00 and 12 UTCVaried weather through the periodNested in operational NAE (12 km)

Testing strategy


Case study 11/08/04 18Z Organised convection

■ Banded structure in the 6hr ppn accumulation compares well with radar.


Case study 11/08/04 18Z Organised convection

■ However the banded structure persists in the 24hr accumulation while the radar image is smooth. ■ Better agreement when averaged to a 12km grid.



“Borders Grid Point Storm” (03/08/2004 00Z, T+13)


UK4 Model Pre-operational tests summary

Very robust Small structure of precipitation too persistent Unrealistically high single row/point precipitation rates (grid point storms) Better forecast of fog than coarser resolution operational models Negative cloud bias (operational models use a cloud enhancement scheme) Diurnal cycle in screen temperature bias due to cloud bias PMSL errors strongly linked to errors in the driving model


Introduction into operations of the UK4 Model

U4.01 13th April 2005 One forecast per day to T+36 at 00Z

U4.02 17th June 2005 Extra T+36 forecast at 12Z added

No operational failures.Negative cloud bias confirmed.Feedback from forecasters:

Positive overall. Number of false alarms of heavy precipitation events. Excessive precipitation rates and organisation of ppn.

Too much light precipitation was observed, as statistics for longer periods were available.


UK4 Operational (North Yorkshire floods)

Rye Valley floods. 19/06/2005

Good guidance at T+17 in intensities, time and locationBut false alarm over South Wales at T+9

12-18Z

12-18Z

12-18Z

Radar 12Z 6 hr rainfall

MES 12Z 6 hr rainfall

4 km 12Z avg 6 hr rainfall E

rror

sca

le (k

m)

Rainfall threshold (mm)

19 June 2005Flash flooding caused by thunderstorms overNorth Yorkshire


UK4 model cycle U4.03 – 12th Oct 2005

Upgrade of convection scheme to current operational version. Keeping the CAPE dependent CAPE closure modification.Substitute horizontal diffusion of moisture by targeted diffusion of moisture.Include cloud enhancement scheme.Tune microphysics parameters to reduce excessive light precipitation


UK4 Operational (Next upgrade)


Planned UK4 changes

December 2005:Introduce 3D VAR assimilation system (plus Analysis Correction scheme for cloud and precipitation).Introduce orographic wave drag parametrisation.Introduce anthropogenic urban heat source.Introduce daily initialisation of soil moisture2006:Increase vertical resolution.Introduce seasonal variability in Leaf Area IndexSplit urban tile type in rooftops and street canyonsReview convection scheme


Questions?

unified model developments 2005 for ewglam/srnwp annual meeting 2005

Documents

uk4 operational

z radar

hr rainfall mes

operational implementation

node sx6

operational models

t3e performance

increase performance