© Crown copyright 2007
Optimal distribution of polar-orbiting sounding missions
John Eyre Met Office, UK CGMS-40; Lugano, Switzerland;5-9 Nov 2012
© Crown copyright 2007
Optimal distribution of polar-orbiting sounding missions
• Background
• Previous studies in Europe
• A new theoretical study:
• the impact of temporal spacing of observations on analysis accuracy
• Conclusions
© Crown copyright 2007
Acknowledgements
Enza Di Tomaso (ECMWF) Niels Bormann (ECMWF)
Peter Weston (Met Office)
© Crown copyright 2007
LST:
00h
LST:
12h
Background:WMO Vision for the GOS in 2025
Recommended baseline with in-orbit redundancy
approved by WMO-EC, 2009
© Crown copyright 2007
Previous studies in Europe
Assimilation of ATOVS radiances at ECMWF.E. Di Tomaso and N.Bormann. EUMETSAT/ECMWF Fellowship Programme Res. Rep. 22
Also presented in CGMS-38 EUM-WP-41
© Crown copyright 2007
Orbits of current satellites
MetOp-A
NOAA-18 NOAA-19
NOAA-15
LECT
© Crown copyright 2007
Data coverage
“NOAA-15 experiment”* MetOp-A * NOAA-18 * NOAA-15
“NOAA-19 experiment”* MetOp-A * NOAA-18 * NOAA-19
Sample coverage from a 6-hour period around 00Z
© Crown copyright 2007
Forecast impact of ATOVS
“ Averaged over extra-Tropics, impact of NOAA-15 experiment versus NOAA-19 experiment is neutral to slightly positive ”
“NOAA-15 exp” RMSE – “NOAA-19 exp” RMSE
“NOAA-19 experiment”
GOOD
“NOAA-15 experiment”
GOOD
Note: AIRS and IASI not assimilated in these experiments
© Crown copyright 2007
New theoretical study: the impact of temporal spacing of observations on analysis accuracy
© Crown copyright 2007
Outline of theoretical study
• Very simple DA system
• one variable in space
• observations distributed in time
• Observations inserted in 12-hour cycle
• to simulate 1, 2, 3 or 4 satellites
• with temporal spacing to simulate 3 orbital planes
• Results found to be very sensitive to assumed rates of forecast error growth
• different rates of doubling time for forecast error variance used:
• 12 hours, 6 hours, 3 hours
• See CGMS-40 WMO-WP-19 for theory and details
© Crown copyright 2007
The experiments:different numbers of observations and
different observation spacings
relative observation time (hours) 0 1 2 3 4 5 6 7 8 9 10 11
experiment number
number of observations
constellation code
1 1 [1,0,0] 1
2 2 [2,0,0] 2
3 2 [1,1,0] 1 1
4 3 [3,0,0] 3
5 3 [2,1,0] 2 1
6 3 [1,2,0] 1 2
7 3 [1,1,1] 1 1 1
8 4 [2,2,0] 2 2
9 4 [1,2,1] 1 2 1
© Crown copyright 2007
Average analysis error variance:forecast error variance doubling time = 12 hours
© Crown copyright 2007
Average analysis error variance:forecast error variance doubling time = 6 hours
note change of scale
© Crown copyright 2007
Average analysis error variance:forecast error variance doubling time = 3 hours
note change of scale
© Crown copyright 2007
For 3-satellite constellations:percentage increases in analysis error variance relative to [1,1,1]
© Crown copyright 2007
For 4-satellite constellations:percentage increases in analysis error: [2,2,0] relative to [1,2,1]
© Crown copyright 2007
Forecast sensitivity to observations (FSO) in global NWP: (Joo, Eyre and Marriott. Met Office FR Tech. Rep. No.562, 2012. Also submitted to MWR.)
Pecentage Contribution of Observations Impacts
0 5 10 15 20 25 30
Ob
serv
atio
n T
ypes
Percentage of Total Observation Impact[%]
Metop
NOAA
"SONDE"
OTHER LEO
AIRCRAFT
SFC LAND
GEO
SFC SEA
OTHER RO
Relevance of theoretical results to real world?
~64% of impact comes from satellite observations
…
of which ~90% from polar sounding data
…
higher for mid-latitude oceans
© Crown copyright 2007
Theoretical study – Conclusions
• Mean analysis error variance is most relevant metric when assessing impact of temporal spacing of observations on global NWP performance
• Dependence of mean analysis error variance on observation spacing is very sensitive to assumed rate of forecast error growth:
• for a 12-hour doubling time of forecast error variance, dependence on observation spacing is significant but small,
• for a 3-hour doubling time reaching ~25% increase in variance for plausible 3-satellite constellations, and ~8% increase for 4-satellite constellations.
• These simple experiments are relevant to real NWP systems, particularly for rapidly-developing storms over mid-latitude oceans.
• Results support assumptions guiding the WMO Vision: that polar-orbiting satellites should be equally space in time, as far as is practicable.
© Crown copyright 2007
Overall conclusions
© Crown copyright 2007
Overall conclusions
• OSE and theoretical study results support guidance that observations should be roughly equally spaced in time
• Impact of observation spacing on NWP is greatest when forecast error growth rates are high, as likely in rapidly-developing storms
• At least one set of IR+MW sounding instruments in an early morning orbit is highly desirable
• More important to optimise the temporal spacing than to hit specific absolute LECTs
© Crown copyright 2007
Thank you! Questions?
© Crown copyright 2007
O
B
A
O
B
A
© Crown copyright 2007
experiment number1 2 3 4 5 6 7 8 9
number of observations1 2 2 3 3 3 3 4 4
Δt constellation code [1,0,0] [2,0,0] [1,1,0] [3,0,0] [2,1,0] [1,2,0] [1,1,1] [2,2,0] [1,2,1]
12 h
mean accuracy 1.485 2.970 2.970 4.454 4.454 4.454 4.454 5.939 5.939
mean error variance 0.701 0.350 0.341 0.234 0.228 0.228 0.225 0.171 0.169
% difference - +2.7 0 +3.6 +1.3 +1.1 0 +0.7 0
6 h
mean accuracy 0.764 1.528 1.528 2.291 2.291 2.291 2.291 3.055 3.055
mean error variance 1.531 0.766 0.690 0.510 0.468 0.462 0.444 0.345 0.336
% difference - +10.9 0 +15.0 +5.5 +4.1 0 +2.9 0
3 h
mean accuracy 0.404 0.808 0.808 1.212 1.212 1.212 1.212 1.616 1.616
mean error variance 4.509 2.254 1.546 1.503 1.108 1.018 0.882 0.773 0.680
% difference - +45.8 0 +70.5 +25.7 +15.5 0 +13.7 0
The experiments:mean analysis accuracies and error variances
© Crown copyright 2007
Forecast sensitivity to observations (FSO): importance of Metop data
Pecentage Contribution of Observations Impacts
0 5 10 15 20 25 30
Ob
serv
atio
n T
ypes
Percentage of Total Observation Impact[%]
Metop
NOAA
"SONDE"
OTHER LEO
AIRCRAFT
SFC LAND
GEO
SFC SEA
OTHER RO
© Crown copyright 2007
Forecast sensitivity to observations (FSO): importance of Metop data
Percentage Contribution of Satellite Impacts (per Platform)
0 10 20 30 40
Ob
serv
atio
n T
ypes
Percentage of total satellite observation impact[%]
MetopNOAA19AquaNOAA15NOAA18MeteosatCOSMICMTSATGOESCoriolisTerraF16NOAA17GRACENOAA16ERS2
© Crown copyright 2007
Forecast sensitivity to observations (FSO): importance of Metop data
Satellite Impacts on NWP forecast(AMVs)
-5-4-3-2-10
Ob
serv
atio
n T
ypes
Observation Impact[J/kg]
GOES13GOES11MTSATMeteosat9Meteosat7TERRAAQUANOAA19NOAA18NOAA17NOAA16NOAA15
Satellite Impacts on NWP Forecast(AMVs/Per sounding)
-1.6E-05-1.4E-05-1.2E-05-0.00001-8E-06-6E-06-4E-06-2E-060
Ob
serv
atio
n T
ypes
Mean Observation Impact[J/kg]
GOES13GOES11MTSATMeteosat9Meteosat7TERRAAQUANOAA19NOAA18NOAA17NOAA16NOAA15
a) b)
Percentage Contribution of Satellite Impacts (per Technique)
0 10 20 30 40 50
Ob
serv
atio
n T
ypes
Percentage of total satellite observation impact[%]
MWS
IRS
Imager/AMV
SCAT
GPSRO
MWI
Mean Satellite Impacts on NWP Forecast(per Technique)
-0.0002-0.00015-0.0001-0.000050
Ob
serv
atio
n T
ypes
Mean Observation Impact[J/kg]
MWS
IRS
Imager/AMV
SCAT
GPSRO
MWI
a) b)
© Crown copyright 2007
Forecast sensitivity to observations (FSO): importance of Metop data
Satellite Impacts on NWP Forecast(Sounders)
-40-30-20-100
Obs
erva
tion
Type
s
Observation Impact[J/kg]
Metop-A/IASIEOS-Aqua/AIRSMetop-A/AMSU-ANOAA19/AMSU-ANOAA18.AMSU-ANOAA15/AMSU-AMetop-A/MHSNOAA18/MHSMetop-A/HIRSNOAA19/HIRSNOAA17/HIRS
Satellite Impacts on NWP Forecast(Sounders)
-6.E-05-5.E-05-4.E-05-3.E-05-2.E-05-1.E-050.E+00
Obs
erva
tion
Typ
es
Mean Observation Impact[J/kg]
Metop-A/IASIEOS-Aqua/AIRSMetop-A/AMSU-ANOAA19/AMSU-ANOAA18.AMSU-ANOAA15/AMSU-AMetop-A/MHSNOAA18/MHSMetop-A/HIRSNOAA19/HIRSNOAA17/HIRS
a) b)
Percentage Contribution of Satellite Impacts(per Metop Sensor)
0 10 20 30 40 50
Obs
erva
tion
Type
s
Percentage of total Metop observation impact[%]
IASI
AMSU-A
ASCAT
MHS
HIRS
GRAS
Mean Satellite Impacts on NWP Forecast(per Metop Sensor)
-0.00012-0.0001-0.00008-0.00006-0.00004-0.000020
Obs
erva
tion
Type
s
Mean Observation Impact[J/kg]
IASI
AMSU-A
ASCAT
MHS
HIRS
GRAS
a) b)