Towards Rapid Update Cycling for Short Range NWP Forecasts in the HIRLAM Community

WMO/WWRP Workshop on Use of NWP for NowcastingUCAR Center Green Campus, Boulder, Colorado, USA

24-26 October, 2011

Magnus Lindskog, Siebren de Haan, Sibbo van der Veen, Sigurdur Thorsteinsson, Shiyu Zhuang, Tomas Landelius and Kristian Pagh Nielsen

Structure

•The HIRLAM consortium

•Developments towards Rapid Update Cycling

•Experimental results

•Concluding remarks

ALADIN/HIRLAM close collaboration started in 2005

Lithuania

Model domains in HIRLAM consortia

DMI HARMONIE (AROME) DMI HARMONIE (AROME) (2.5 km hor res, 65 vert lev)(2.5 km hor res, 65 vert lev)

HIRLAM 7.4 RCR HIRLAM 7.4 RCR (7 km hor res, 65 vert lev)(7 km hor res, 65 vert lev)

HIRLAM 7.3 RCR HIRLAM 7.3 RCR (15 km hor res, 60 vert lev)(15 km hor res, 60 vert lev)

SMHI HARMONIE (ALARO) SMHI HARMONIE (ALARO) (5.5 km hor res, 60 vert lev)(5.5 km hor res, 60 vert lev)

HIRLAM ref DA: 4D-VarHIRLAM ref DA: 4D-VarHARMONIE ref DA: 3D-VarHARMONIE ref DA: 3D-Var

Focus is moving towards frequently updated short-range km-scale forecastsFocus is moving towards frequently updated short-range km-scale forecasts

Towards Rapid Update Cycling (RUC)On-going data assimilation developments

•Investigate effects of increasing frequency of data assimilation cycles and of shortening observation cut-off time

•Utilization of new types of observations

•Handling of balances

•Algorithmic developments

New types of observations

• RADAR radial winds and reflectivities, GNSS (GPS) ZTD, Mode-S, satellite based radiances (IASI, SEVIRI,ATOVS), GPS RO, derived satellite based cloud-products, Scatterometer,…

•ASCAT, SMOS, MODIS, GLOBSNOW, …

Upper-air:

Surface:

Illustration derived cloud productsIllustration upper-air observation types

Handling of Balances

(seasonal variation of coupling of humidity background errors with errors of other variables as derived for km-scale model over Danish domain)

Air-mass/flow dependence to be represented

SUMMER (12 UTC)

VORTICITYDIVERG.T and Ps

WINTER ( 12UTC)

Algorithmic developments

•EKF for surface DA

•4D-Var

•ETKF, EnDA

•Hybrid DA

•Handling of non-additive errors

Estimate the phase error (displacement field) and warp the background state.

Minimize the additive error using standard VAR-method.

Non-additive errors

bHx

y

)(sxb

)( Tsxb T

Estimate Warp

Handling – two step method

(phase-/displacement-/alignement-/timing errors)

H(fg) Estimated T SEVIRI

Example

Domains D11/H11/U11: 11 km hor res., 60 vert lev.

HIRLAM model RUC parallel experiments

Summer period: 1 May 2010- 5 September 2010 Winter period: 13 January 2011- 28 February 2011

Parallel experiments over H11 and U11 domains

Experimental Design

Verification of Rainfall forecasts

Verification area

siebren.de.haan@knmi.nl

1. Transfer of MSG cloud cover to 3D cloud cover in HIRLAM model:

•cloud cover N from NWC SAF•cloud base from (interpolated) synoptic observations•cloud top from MSG (10.8 micron channel)

2. Translate N to humidity

Parallel experiments U11 RUC with and without cloud initialization

Verification results by comparison of Hirlam cloudiness to synoptic observations (bias and standard deviation of errors)(large verification area over Europe)REF: Hirlam reference run MSG: Hirlam run with MSG cloud initialisation

Cloud forecast Verification scores

HARMONIE system parallel experiments

• 6 h intermittent data assimilation cycle

• 3 h intermittent data assimilation cycle

Two parallel exp. for July & August 2009 and January & February. 2010:

Model domain:Model domain: SMHI pre-oper. SMHI pre-oper.Horizontal resolution:Horizontal resolution: 5.5 km 5.5 kmVertical levels: Vertical levels: 6060LBC:LBC: 3 hourly with ECMWF fc 3 hourly with ECMWF fcSurface DA:Surface DA: Optimal Interpolation Optimal InterpolationUpper–air DA:Upper–air DA: 3D-Var 3D-VarObservation usage:Observation usage: SYNOP, SHIP, SYNOP, SHIP, DRIBU, TEMP, PILOT, AIREP, AMDAR, DRIBU, TEMP, PILOT, AIREP, AMDAR, Conv.+ATOVS AMSU-A Conv.+ATOVS AMSU-A Initialization:Initialization: IDFI IDFI

Scores for verification against observations

(summer period)

6h cycle

Surface pressure (hPa) RMS/BIAS as function of forecast

range

Temperature (K) RMS/BIAS of + 12 h forecasts as function of vertical level

3h cycle

Assimilation of ATOVS AMSU-A crucial for positive impact of 3h data assimilation cycle in this parallel experiment

Conclusions and Future Plans•Utilization of observations with high resolution in space and time important for RUC.

•Encouraging first results from initializing clouds for RUC, applying a simple approach.

•Significant seasonal variations of balances revealed for a km-scale model. Future plans include investigation of air-mass and flow dependent balances. Imbalances and associated spin-up need further investigations.

•Algorithmic developments for handling of non-linearities, complex observation types and non-additive errors are on-going.

•Co-ordinated impact studies planned to assess the impact of new observation types and to optimize the handling of these.