LACE Data Assimilation Working Days 14-16 June 2011, Budapest HARMONIE DA suite at KNMI Jan Barkmeijer With contributions from: Cisco de Bruijn, Siebren.

LACE Data AssimilationWorking Days

14-16 June 2011, Budapest

HARMONIE DA suiteat KNMI

Jan Barkmeijer

With contributions from: Cisco de Bruijn, Siebren de Haan, Gert-Jan Marseille, Emiel van der Plas, and Wim de Rooy.

Content

- Description KNMI DA suite

- Monitoring tools

- Problems encountered

- Future developments

3

The KNMI Data assimilation suite

• Initial run at 15 November 2010, but many changes (experimental status) since then.

• Runs at ECMWF, using hourly boundaries and observation set from a H7.2 Hirlam suite at KNMI (ectrans is used to tranfer data vice-versa)

• Cycling time is 6h and a 24h forecast is produced at 00 and 12 UTC.

• Integration area (300x300 gp at 2.5km grid resolution) covers the Netherlands and vertical resolution is 60 levels (Meteo-France definition). Lowest model level around 9m

• Harmonie model cycle H36h1.2 with some modifications

• Data is processed e.g. at DMI/KNMI for monitoring purposes.

Area definition

200x200

300x300

400x400

500x500

LBC data from

Hirlam forecast

for both runs

300x300 400x400

RADAR

Computation of background error statistics

- Used the recipe as described on the Hirlam wiki page for Harmonie Data-assimilation

- Central here is the creation of a series of downscaled Harmonie 6h forecasts for 20060920-20061031 at 00 UTC, using 4 members of the ECMWF EPS

Confusion w.r.t. virtual and real temperature (LSPRT)

DMI Harmonie suite KNMI old

Courtesy Nils Gustafsson

After correction

DMI KNMI new

Content

- Description KNMI DA suite

- Monitoring tools

- Problems encountered

- Future developments

What datasets are being used?

SYNOP U,V Z T

AIREP U,V T

RADIOSONDE U,V Z T Q

Mode-S U,V T

What data sets are being used (2)?

Mode-S drops out

o-a/o-b statistics for April

negative bias

The Danmark Monotoring Facility

Monitoring with KNMI Parametrization Testbed (KPT)

1D column at Cabauw from 3D run

Occurrence of fog

Occurrence of fog (2)

Content

- Description KNMI DA suite

- Monitoring tools

- Problems encountered

- Future developments

Modis VIS satellite image13th of May 2008 1315 UTC Cloud cover Cloud cover

EDMFM and cloud scheme changes

Increase variance computation in the cloud scheme to alleviate the on-offcharacter in the generation of cloud

Instabilities in the canopy model over sea (ECUME)

Using Meteo-France 60 level definition

Switching off ECUME over sea

Occurrence of negative q in the analysisAlso other hydrometeors are affected (e.g. cloud-ice and cloud-water)

First-guess Analysis

Solution: set GFL >=0 at analysis time

Impact on 24-h accumulated precipitation

original analysis modified

Vertical nesting strategy 1.3hPa 10hPa 0.01hPa

Vertical nesting strategy 1.3hPa 10hPa 0.01hPa

Since 1 June 2011 a parallel 3dvar suite runs atECMWF (cy36h1.3) and with ECMWF boundaries

T+12h forecast for T2m valid 5 June 2011 at 12 UTC

HIRLAM LBC ECMWF LBC

RADAR 3DVAR + HIRLAM LBC

400x400 500x500

RADAR 3DVAR + ECMWF LBC

400x400 500x500

FUTURE PLANS- Get radar data-assimilation running

- Implementation of a 800x800xL60 Harmonie suite at KNMI, 3dvar with cycling time of 3 hour (or less)

- Add more q observations: GPS, Cloud data (e.g. cover)

- Develop more tools for monitoring (MetEvaluation)

- Coupling with other applications:surge model, LOTOS (air pollution model)

- Increase research effort on DA+EPS:+ timing/location of convection+ 4dvar without nonlinear updates