Assimilation of GPS radio-occultation observations at Meteo-France

Nathalie Saint-Ramond

With contributions from Pierre Brousseau CNRM-GAME, Météo-France and CNRS

ICGPSRO 14-16 May 2013, Taïwan.

Outline

2



GPS RO operational assimilation



Change in the vertical thinning



Test of GPSRO in the non-hydrostatic limited area model AROME



Summary and Outlook

GPS RO operational assimilation

3

GPS RO operational assimilation Assimilated in the global 4DVAR : ARPEGE (collaboration with ECMWF) and LAM 3DVAR : ALADIN since september 2007



Bending angles



Rising and setting occultations Up to 36 km



Accounting for tangent point drift



Screening



Horizontal / Vertical thinning 1 datum per model vertical layer

T798 C2.4 L70

4

GPS RO operational assimilation Local resolution (in km) in T798 C2.4

Guiana Polynesia

Reunion New Caledonia

Highest resolution : T1920, e.g. 10 km over France Lowest resolution : T333, e. g. 60 km on the opposite side of the Globe ARPEGE – ALADIN : same vertical levels 5

GPS RO operational assimilation

Height in km

ARPEGE 70 levels ECMWF 91 levels

36

Layer thickness in m 6

GPS RO operational assimilation DFS

Observation number

0.82% Sat 88%

Sat 68%

6.6%

AIRS AMSU-A GPSRO

IASI AIRCRAFTS ground 12%

1 day: 14 August 2012 7

TEMP

ground 32%

Forecast Sensitivity to Observation Why ?

Such a large number of observations needs to be monitored to indicate their influence on forecasts skills 8

Forecast Sensitivity to Observation How ? 

Linear estimate of observation impact x bf

J



J : 3D integrated dry total energy of the difference between the 24h forecast and a reference state



Observation impact

xaf

time t-1

t0

a 1 1  T J b T J  deltaJ  ( R HA) M a f  M b f ( y  Hxb ) 2 xb xa  

• second order approximation (Errico, 2007).

9

deltaJ

t1

GPSRO operational assimilation Impact on the 24h forecast

height (km)

Impact depending on height

10

GPSRO operational assimilation Impact on the 24h forecast

Impact depending on platform

height (km)

Impact depending on height

11

GPSRO operational assimilation Impact on the 24h forecast

Impact per observation Impact depending on platform

height (km)

Impact depending on height

12

Outline

 GPS RO operational assimilation  Change in the vertical thinning  Test of GPSRO in the non-hydrostatic limited area model AROME  Summary and Outlook

13

GPSRO vertical thinning Fraction of used observations:

40000

not used used 37000

Only 1 obs per model level were used

14

GPSRO vertical thinning Number of available observations: More available observations in the test model due to the addition of Metop-B data

Number of used observations: More observations used due to Metop-B and the vertical thinning modification 5 times more observations used

15

TEST OPER

GPSRO vertical thinning departure statistics of the bending angles RMS

BIAS

TEST/OPER (o-b)/b (o-a)/a

(hPa)

% Scores for the model T wrt ECMWF and radiosondes:

%

TEST/OPER

T / RS 00H

T/ECMWF Analyses 00H

RMS BIAS

16

(K)

(K)

GPSRO impact on 24h forecast With vertical thinning : OPER

Impact depending on height

17

Without vertical thinning : TEST

Other modifications in the experimental model 

New observations:  CrIs: Infrared sounder on Suomi NPP  ATMS: microwave sounder on Suomi NPP  Metop-B instruments :    

IASI (more WV channels, observation error reduced) AMSU-A MHS GRAS

 OSCAT on OceanSat-2: good impact on cyclone forecast over « La Réunion » (observation error aslo modified)  Radiances from GOES-13 and GOES-15: 1 high tropospheric channel  More channels from MHS on NOAA-19  More ground GPS



Observation errors in the models were tuned in order to balance the minimisation, using Desroziers’s diagnosis: modified by a factor  0.75 for SYNOP, AIRCRAFTS, SATWIND, BUOYS, RS, PILOT  1.35 for GPS RO

18

Outline

 GPS RO operational assimilation  Change in the vertical thinning  Test of GPSRO in the non-hydrostatic limited area model AROME  Summary and Outlook

19

Tests in the non-hydrostatic model AROME Model description: -2.5 km horizontal resolution

-60 levels -Non-hydrostatic -3DVAR assimilation system : -Same data available as for ARPEGE -Fine scale data from radar network ARPEGE 10km

20

AROME 2.5 km

Tests in the non-hydrostatic model AROME Results on rain 24h forecasts: Probability Of Detection OPER TEST 0.5mm

2mm

5mm

10mm

False Alarm Rate OPER TEST

10mm 0.5mm

21

2mm

5mm

Summary and Outlook  GPS RO operationaly assimilated since 2007 in global and LA Models 1D + AD/TL from GRAS-SAF  Good impact on upper troposphere / lower stratosphere but also at higher levels where few other observations are assimilated.  Remove of vertical thinnning has proven to have a good impact on model performences with observation errors adjusted  Impact in the non-hydrostatic model AROME is not clear for the moment. 22

Thank you