Kick-Off-Treffen SPP, Bonn 23-24 October 2006 Improved Water Vapour and Wind Initialisation for Precipitation Forecasts: Impact Studies with the ECMWF.

Kick-Off-Treffen SPP, Bonn 23-24 October 2006

Improved Water Vapour and Wind Initialisation for Precipitation Forecasts: Impact Studies with the ECMWF IFS

Martin Weissmann, Andreas Dörnbrack, Gerhard Ehret, Christoph Kiemle, Stephan Rahm, Oliver Reitebuch

Institut für Physik der Atmosphäre, DLR Oberpfaffenhofen, Germany

Carla Cardinali, Elias HolmEuropean Centre for Medium-Range Weather Forecasts (ECMWF), Reading, UK


Motivation

above oceans:hardly any radiosondesaircraft at cruise level height errors of AMVslow accuracy of passive instruments low vertical and horizontal resolution

lidars can measure various atmospheric quantities in remoteregions with high accuracy and highresolution either from satellites or aircraft

goal: estimate benefit with impact studies

AMSU-A:50-55 GHz


DLR lidar instruments

Doppler lidardropsondes,u, v, t, rh, p

scanning coherent 2 µm Doppler lidarconical scans with 24 azimuth positions vertical profile of 3-D wind vectorhoriz. resolution 5 - 40 kmvert. resolution 100 m

Differential Absorption Lidar (DIAL) nadir or zenith pointingprofile of water vapour molecule numberhoriz. resolution: 2 - 40 kmvert. resolution: 500 - 2000 m

DIAL lidar


4 flights in "sensitive areas" (targeting)1 flight for Greenland Tip Jet1 flight for intercomparison ASAR and lidar2 transfer flights=======================================8 flights, 1600 wind profiles, 40 000 lidar measurements, 49 dropsondes

Wind data - Atlantic THORPEX Regional Campaign (A-TReC)

5 million operational measurements used per daylidar = 0.005% additional measurements

DLR Falcon:14 - 28 Nov. 2003


Iceland

Ireland

ECMWF sensitivity plot and 500 hPa

http://www.sat.dundee.ac.uk/

Observations on 25 November 2003


Dropsonde(2000 times enlarged)

Error lidar1-1.5 m/s

Error drops/raso2-3 m/s

Instrumental and representativeness errors

Error AMV2-5 m/s

Background departures:(Std(bg-dep))2 = (Stdobs)2 + (Stdbg)2


observation influence (Cardinali et al. 2004):0 --> no influence of observations1 --> no influence of backgroundmean global observation influence = 0.15

Observation influence (22 November 2003)


(gpd

m)

Reduction of forecast error at 500 hPa - 72 h

area: 17 x 10-6 km2


Mean reduction over Europe, averaged over 29 forecasts (~2 weeks)black: experiments with lidar, gray: experiment with 100 dropsondes

Reduction of forecast error - 500 hPa


Reduction of forecast error of 500 hPa geopotential height:Lidar 72 h: ~ 1 m (3.5%)

Simmons and Hollingsworth 2002:72 h: 10 m in 10 years

Comparison to mean reduction of NWP error


Reduction of forecast error - 48, 72, 96 h


Reduction of u, v, z, rh, and t forecast errors

Relative reduction of RH forecast error - 48, 72, 96 h


NOAA

DLR

20031115 at 18 Step42 20031120 at 18 Step54

20031122 at 18 Step66 20031125 at 18 Step30

Cut off low; heavy precipitation in SE-Spain

Heavy rain/strong wind in Portugaland NW-Spain

Strong winds –northwest Europe Heavy rain, Scotland and Norway

A-TReC targeting campaigns


The influence of targeted observations

-1.2

-0.8

-0.4

0

0.4

0.8

1.2

1.6

Airep Lidar Temp Drop

ATReC 200112018

Contribution to FCE Obs Rel Num

-1.2

-0.8

-0.4

0

0.4

0.8

1.2

1.6


ATReC 2003112218

Contribution to FCE ObsRel Num

-1.2

-0.8

-0.4

0

0.4

0.8

1.2

1.6


ATRec 2003112518


-1.2

-0.8

-0.4

0

0.4

0.8

1.2

1.6


ATReC 2003111518


+

+ statistical problem --> larger sample63 - 90% of total impact outside of verification area --> larger areasimproved sensitivity predictionsbetter structure functionspractical restrictions --> better planning


DIAL lidar water vapour observations

Transfer flights to 4 field campaigns during 2002 - 2005


Impact of lidar H2O-observations (preliminary)

Longitude

La

titu

de

60 W 20 E20 W

Reduction of TCWV first guess error (lidar - control)

WV

mixing ratio [m

g/kg]


Conclusions

first assimilation of "real" Doppler lidar measurements in global NWP model lidar measurements have a smaller error than all other operational wind observations --> high analysis influence and information content

lidar wind measurements reduce the average forecast error of u, v, z, rh, and t over Europeaverage reduction of the 48 - 96h forecast error over Europe ~3%

limitations of targeted observations:need for more cases, larger verification area, systematic decisions, better structure functions...

ongoing water vapour studies

emphasizes the potential of airborne and spaceborne lidars (ADM-Aeolus, possible future water vapour DIAL satellite)

future campaigns: COPS 2007, IPY 2008, T-PARC 2008

www.pa.op.dlr.de/na-trec/

Kick-Off-Treffen SPP, Bonn 23-24 October 2006 Improved Water Vapour and Wind Initialisation for Precipitation Forecasts: Impact Studies with the ECMWF.

Documents