Lessons learned from THORPEX Lessons learned from THORPEX
THORPEX working group on Data THORPEX working group on Data Assimilation and Observing StrategiesAssimilation and Observing Strategies
Florence Rabier (Météo-France and CNRS, France, Co-chair)Pierre Gauthier (UQAM, Canada,Co-chair) Carla Cardinali (ECMWF, Int)Ron Gelaro (GMAO, USA) Ko Koizumi (JMA, Japan) Rolf Langland (NRL, USA)Andrew Lorenc (Met Office, UK)Peter Steinle (BMRC, Australia)Mickael Tsyrulnikov (HRCR, Russia)
Nonlinear Processes in Geophysics, 15, 1-14, 2008
New WG being formed, including Observing Systems
THORPEX and the DAOS-WGTHORPEX and the DAOS-WG
• ““THORPEXTHORPEX: a Global Atmospheric Research Programme” established in : a Global Atmospheric Research Programme” established in 2003 by WMO.2003 by WMO.
• Mission statementMission statement: “Accelerating improvements in the accuracy of high-: “Accelerating improvements in the accuracy of high-impact 1-14 day weather forecasts for the benefit of society and the economy”impact 1-14 day weather forecasts for the benefit of society and the economy”
• Design and demonstration of Design and demonstration of interactive forecast systemsinteractive forecast systems: enhancements to : enhancements to the observations usage in “sensitive regions”the observations usage in “sensitive regions”
• Perform THORPEX Perform THORPEX Observing-System Tests and Regional field CampaignsObserving-System Tests and Regional field Campaigns to test and evaluate experimental remote-sensing and in-situ observing to test and evaluate experimental remote-sensing and in-situ observing systemssystems
• DAOS-WGDAOS-WG: evaluate and improve the impact of observations : evaluate and improve the impact of observations
OutlineOutline
• ContextContext• Main objectivesMain objectives
– Assess impact of observations and observing system design
– Targeting strategies– Improved use of observations
• Illustrations from field campaigns (AMMA…), Illustrations from field campaigns (AMMA…),
the Intercomparison experiment and the WMO the Intercomparison experiment and the WMO Data Impact WorkshopData Impact Workshop(http://www.wmo.int/pages/prog/www/OSY/Reports/NWP-4_Geneva2008_index.html)(http://www.wmo.int/pages/prog/www/OSY/Reports/NWP-4_Geneva2008_index.html)
Large number of data and different data sources
Assessing the impact of Assessing the impact of observationsobservations
• OSEsOSEs
• OSSEsOSSEs
• DFSDFS
• Error variance reductionError variance reduction
• Sensitivity to observationsSensitivity to observations
Winter results: Baseline – Control (Z500)Winter results: Baseline – Control (Z500)Impact of terrestrial, non-climate, observationsImpact of terrestrial, non-climate, observations
NH
EUR
Differences in RMS errors and significance bars for each forecast range
ECMWF
Control-Baseline (Z500)Control-Baseline (Z500)Normalised forecast error difference, Day-3Normalised forecast error difference, Day-3
Geographical distribution of error reduction ECMWF
Neutral Case impact A few hours 6 hours 12 hours
NorthernHemisphereExtra-tropics
Radiosonde
Aircraft
Buoys
AIRS
IASI
AMSU/A
GPS-RO
SCAT
AMV
SSMI
Tropics
Radiosonde
Aircraft
Buoys
AIRS
IASI
AMSU/A
GPS-RO
SCAT
AMV
SSMI
SouthernHemisphereExtra-tropics
Radiosonde
Aircraft
Buoys
AIRS
IASI
AMSU/A
GPSRO
SCAT
AMV
SSMI
Synthesis of all results after WMO workshop
Analysis
Nature run(output from high
resolution, high qualityclimate model)
Simulator
Forecastmodel
Candidateobservations
(e.g. GEO MW)Initial conditions
Referenceobservations
(RAOB, TOVS,GEO, surface,aircraft, etc.) Forecast
products
Assessment
OSSE, conceptual model
End products
JCSDA
Vertical structure of a HL vortex shows distinct eye-like feature and prominent warm core; low-level wind speeds exceed 55 m/s
Reale O., J. Terry, M. Masutani, E. Andersson, L. P. Riishojgaard, J. C. Jusem (2007), Preliminary evaluation of the European Centre for Medium-Range Weather Forecasts' (ECMWF) Nature Run over the tropical Atlantic and African monsoon region, Geophys. Res. Lett., 34, L22810, doi:10.1029/2007GL031640.
HL vortices: vertical structureHL vortices: vertical structure
Tropical cyclone NR validation
Preliminary findings suggest good degree of realism of Atlantic tropical cyclones in ECMWF NR.
DFS: Information content by areaDFS: Information content by area
M-F
DFS= Tr(HK)=Tr(I-AB-1)
Ensemble variational assimilationEnsemble variational assimilationat Météo-Franceat Météo-France
• Ensemble assimilation : simulation of the joint evolution Ensemble assimilation : simulation of the joint evolution of analysis, background and observation errors:of analysis, background and observation errors:
a a = = ((I – KH) I – KH) b b + + K K oo..
• Observations are explicited perturbed, while backgrounds Observations are explicited perturbed, while backgrounds are implicitly perturbed through cycling.are implicitly perturbed through cycling.
(From Ehrendorfer, 2006)
Ensemble Ensemble bb – – a a with with energy normenergy norm
One month statistics (January 2007) at 00UTC
6 member 3D-Var FGAT ensembleDesroziers, M-F
14
Observations move the model state from the “background” trajectory to the new “analysis” trajectory
The difference in forecast error norms, , is due to the combined impact of all observations assimilated at 00UTC
Sensitivity to Observation Sensitivity to Observation ((Langland and Baker, 2004)Langland and Baker, 2004)
24 30e e
OBSERVATIONS ASSIMILATED
00UTC + 24h
24e30e
Forecast error measure (dry energy, sfc–140 hPa):
Estimating Observation Impact
)()( v0T
v0 xxCxx ffe
gxxx
xxx
x T0
203
0
3
020
2
00 )(...)
6
1
2
1(
eee
e
Taylor expansion of change in due to change in :e 0x
3rd order approximation of in observation space:
)]()([)( vaTavb
Tb
TT xxCMxxCMKy ffe
e
model adjointanalysis adjoint
3T ~)( gy
…summed observation
impact
Analysis equation allows transformation to observation-space:
yKxxx ba0
3T ~)( gy e
Properties of the Impact Estimate
0e0e
…the observation improves the forecast
…the observation degrades the forecast
…see Langland and Baker (2004), Errico (2007), Gelaro et al. (2007)
The “weight” vector is computed only once, and involves the entire set of observations…removing or changing the properties of one observation changes the weight of all other observations.
3~g
Valid forecast range limited by tangent linear assumption for TM
The impact of arbitrary subsets of observations (e.g. instrument type, channel, location) can be easily quantified by summing only the terms involving the desired elements of .y
Forecast error norms and differences
e30
e24
Forecasts from 0600 and 1800 UTC have larger errors
e24 – e30 (nonlinear) e24 – e30 (adjoint)
Global forecast error total energy norm (J kg-1)
Forecast errors on background-trajectories
Forecast errors on analysis-trajectories
NRL
NAVDAS-NOGAPS
Percent of observations that produce forecast error reduction (e24 – e30 < 0)
NRL
AMMA RAOB AMMA RAOB Temperature Ob Temperature Ob Impacts Impacts May-Oct May-Oct
20062006
TAMANASET:60680 SUM= -0.2791 J kg-1
BANAKO:61291 SUM= -0.5755 J kg-1
NRL
Example : AMV impact Example : AMV impact problemproblem
Date: Jan-Feb 2006
Issue: Non-beneficial impact from MTSAT AMVs at edge of coverage area
Action Taken: Data provider identified problem with wind processing algorithm.
NRL
Comparison and Interpretation of ADJ and OSE Results
The ADJ measures the impact of observations in each analysis cycle separately and against the control background, while the OSE measures the impact of removing information from both the background and analysis in a cumulative manner
The ADJ measures the impacts of observations in the context of all other observations present in the assimilation system, while the OSE changes/degrades the system ( differs for each OSE member)
Comparison is restricted to the forecast range and metric for which the adjoint results are valid on the one hand (24h-energy in this study) and to the observing systems tested in the OSE on the other
…a few things to keep in mind…
K
Gelaro
Removal of AMSUA results in large increase in AIRS (and other) impacts
Removal of AIRS results in significant increase in AMSUA impact
Removal of raobs results in significant increase in AMSUA, aircraft and other impacts (but not AIRS)
Combined Use of ADJ and OSEs (Gelaro et al., 2008)
…ADJ applied to various OSE members to examine how the mix of observations influences their impacts
NASA, GMAO
Total observation impact at 00 UTCTotal observation impact at 00 UTCNAVDAS 24h Ob Impact Jan2007 00z+06z
-100 -80 -60 -40 -20 0
AMSUA
Aircraft
LandSfc
MODIS
Windsat
Qscat
RaobDsnd
SSMIspd
SatWind
Ships
Targeting strategiesTargeting strategies
Evaluating and improving targeting strategiesEvaluating and improving targeting strategies
• Select additional observations or optimize the use of satellite sensors (sampling rate, thinning, chanel selection…)
• Results depend on method, flow regimes
• To be extended to Tropics (model error), evaluation at finer scales
Observation time
Adjoint model orEnsemble Transform
Verification time
A-TReC A-TReC ((Atlantic THORPEX Regional Atlantic THORPEX Regional Campaign)Campaign) Oct15-Dec17 2003 Oct15-Dec17 2003
• The ATREC was led by EUCOS in the context of THORPEX. It involved UK Met office, ECMWF, Meteo-France, NRL, NASA, U of North Dakota, Meteorological Service of Canada, NCEP, FSL, NCAR and U of Miami
• A variety of observing platforms were deployed. AMDAR (550), ASAP ships (13), radiosondes (66), GOES rapid-scan winds and dropsondes.
0
10
20
30
40
50
60
70
forecast range (hours)
RM
S (
m)
With 850 hPaWithout 850 hPaWith 500 hPaWithout 500 hPa
Fourrié, et al, M-F
Geopotential forecast error for Geopotential forecast error for the ATReC area the ATReC area
(wrt analyses)(wrt analyses)
Impact of targeted obsImpact of targeted obs
• Targeting is possible and successful – mid-latitude targeted Targeting is possible and successful – mid-latitude targeted observations are about twice as effective as random observations.observations are about twice as effective as random observations.
• Improvements to DA methods should improve the assimilation of all Improvements to DA methods should improve the assimilation of all observations in sensitive regions, including targeted obs, but the observations in sensitive regions, including targeted obs, but the statistical basis still means that only just over 50% will have a positive statistical basis still means that only just over 50% will have a positive impact.impact.
• Improvements to targeting methods are possible (e.g. longer leads for Improvements to targeting methods are possible (e.g. longer leads for large areas) but the statistical basis means that impacts on scores will large areas) but the statistical basis means that impacts on scores will vary.vary.
• Thanks to the general improvement of operational NWP, the average Thanks to the general improvement of operational NWP, the average impact of individual observing systems is decreasing.impact of individual observing systems is decreasing.
• Targeting alone is unlikely to significantly accelerate improvements in Targeting alone is unlikely to significantly accelerate improvements in the accuracy of 1 to 14-day weather forecasts compared to other the accuracy of 1 to 14-day weather forecasts compared to other improvements over the THORPEX period in NWP and satellites.improvements over the THORPEX period in NWP and satellites.
Improving the use of Improving the use of observationsobservations
• Extending the use of satellite dataExtending the use of satellite data
• Bias correctionBias correction
Improved representation of surface emissivity for Improved representation of surface emissivity for the assimilation of microwave observationsthe assimilation of microwave observations
•Dynamical approach for the estimation of the emissivity from Satellite observations over land (Karbou 2006)
•The estimation of emissivity has been adapted to Antarctica : snow and sea ice surfaces
Karbou, M-F
Comparison of the new emissivity calculation with the old Comparison of the new emissivity calculation with the old one, over sea iceone, over sea ice
Fg-departure (K) (obs- first guess) histograms for AMSU-A, ch4 (July 2007)
Fg-departure (K) (obs- first guess) histograms for AMSU-B, ch2 (July 2007)
Old
New
Use of additional microwave dataUse of additional microwave data
AMSUB- Ch3 AMSUA- Ch5
CONTROL
EXP
Density
of data
Being
actively
assimilated
Bouchard, Karbou, M-F
AMMA: The African AMMA: The African Monsoon Monsoon
Multidisciplinary Multidisciplinary AnalysisAnalysis
Better understand the mechanisms of the African monsoon and prevent dramatic situations
(Redelsperger et al, 2006)
Enhanced observations over West Africa in 2006
In particular, major effort to enhance the radiosonde network
(Parker et al, 2008)
Impact of using the AMMA radiosonde Impact of using the AMMA radiosonde datasetdataset
• New radiosonde stationsNew radiosonde stations
• Enhanced time samplingEnhanced time sampling
• AMMA databaseAMMA database: additional : additional data which were not received data which were not received in real time + enhanced vertical in real time + enhanced vertical resolution resolution
• Bias correction for RHBias correction for RH developed at ECMWF developed at ECMWF
(Agusti-Panareda et al) (Agusti-Panareda et al)
• Data impact studies Data impact studies
With various datasets,With various datasets,
With and without RH bias With and without RH bias correctioncorrection
Number of soundings provided on GTS in 2006 and 2005
Period: 15 July- 15 September, 0 and 12 UTC
Impact on quantitative prediction Impact on quantitative prediction of precipitation over Africaof precipitation over Africa
Higher scores for AMMABC
Lowest scores for NO AMMA
CNTR: data from GTS
AMMA: from the AMMA database
AMMABC: AMMA + bias correction
PreAMMA: with a 2005 network
NOAMMA: No Radiosonde data
Faccani et al, M-F
Work performed and lessons learntWork performed and lessons learnt
• Impact of observationsImpact of observations– Guidance for observation campaigns and the configuration of the
Global Observing system– Assessment of the value of targeted observations (papers by Buizza,
Cardinali, Kelly, in QJRMS)– Evaluation of observation impact with different systems (A-TReC,
AMMA…). Need for relevant bias correction.– Intercomparison experiment for sensitivity to observations
• Improving the use of satellite dataImproving the use of satellite data– Extend our use of satellite data (density, cloudy/rainy, over land)
• Important to study different methods and different Important to study different methods and different systems to draw relevant conclusionssystems to draw relevant conclusions