overview of observing system evaluation in godae oceanviewgodae-data/symposium/gov... ·...
TRANSCRIPT
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Overview of Observing System Evaluation in GODAE OceanView
Peter Oke1, Gilles Larnicol2
and all task team members
1CSIRO Marine and Atmospheric Research; 2CLS
November 2013
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
GOV OSEval-TT organisation
Co-Chairs:
Peter Oke (CSIRO)
Gilles Larnicol (CLS)
Associate members:
Mike Bell (UKMet)
Eric Dombrowsky (Mercator-Océan)
Fabrice Hernandez (Mercator-Océan)
Eric Lindstrom (NASA)
Andreas Schiller (CSIRO)
Hans Bonekamp (EUMETSAT)
Core Members:
Magdalena Balmaseda (ECMWF)
Anthony Weaver (CERFACS)
Laurent Bertino (NERSC)
Pavel Sakov (NERSC)
Gary Brassington (BoM)
Jim Cummings (NRL)
Pat Hogan (NRL)
Yosuke Fujii (JMA/MRI)
Villy Kourafalou (Univ. Miami)
Daniel Lea (UKMet)
Matthew Martin (UKMet)
Avichal Mehra (NOAA)
Elisabeth Remy (Mercator-océan)
Greg Smith (EC)
Clementa Tanajura (Remo)
Zhang Tianyu (NMEFC)
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
To provide consistent and scientifically justified
requirements and feedbacks to agencies in charge of
Global and Regional Ocean Observing Systems
GOV OSEval TT Objectives
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Evaluation studies
Perform impact studies of GOOS and ROOS on forecast and reanalysis
GODAE systems
Methodologies: OSE, OSSE and alternative methods
Provision and dissemination of Observation
Impact Statements (OIS) based on OSEval
evidence
Capacity Building
GOV OSEval TT Objectives
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Organisation of workshops to provide a forum for interactions:
• discuss best-practice methods and details of the technical and scientific aspects of impact studies;
• share experiences with new members; and
• coordinate activities and synthesize results in order to provide homogeneous feedback to observational agencies.
Relationships to broader communities:
• Coordination with OOPC and Clivar/GSOP on the climate and seasonal
forecasting applications progressively maintained
Capacity Building
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Capacity Building
1st GODAE OSE/OSSE
workshop
Nov 07 Nov 08 June 09 Oct 10 June 11 Nov 11 Nov 12 Sept 14
(Paris) (Nice) (Toulouse) (Tokyo) (Santa Cruz) (Paris) (Rio) (Europe)
2nd GOV OSEval TT
workshop
OSEval-TT established, definition of a work plan
Consensus on the Routine Monitoring of Obs systems
Emergence of the idea of a TT dedicated to OSE
3rd wks: GOV &
CLIVAR GSOP
NRT OSE set up & demonstration
Concept of “Observation Impact Statement”
GOVST-II GOVST-I GOVST-III GOVST-IV
4th
workshop
To be organised - with E-AIMS European project
Nov 13
(Washington)
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Provide feedback and requirements to the observation agencies
through evidence of impact of the observations in GODAE systems
Examples of responses to “observing system events”: - J1 new orbit (2009): recomm for an orbit that preserves mesoscale sampling
- J1 end of life (2010): recomm to maintain a good altimeter constellation
-ENVISAT loss (2012): collect feedbacks on the impact of the EN loss.
- Cryosat 2 ocean data (2012): recomm to deliver C2 ocean data
- XBT (2012): justification for continuance of the XBT line in Australia
- TAO (2013): GOV OSEval has been asked to provide feedback and requirements
7
Provision and dissemination of
Observation Impact Statements
(OIS)
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Provide feedback and requirements to the observation agencies
through evidence of impact of the observations in GODAE systems
Examples of responses to “observing system events”: - J1 new orbit (2009): recomm for an orbit that preserves mesoscale sampling
- J1 end of life (2010): recomm to maintain a good altimeter constellation
-ENVISAT loss (2012): collect feedbacks on the impact of the EN loss.
- Cryosat 2 ocean data (2012): recomm to deliver C2 ocean data
- XBT (2012): justification for continuance of the XBT line in Australia
- TAO (2013): GOV OSEval has been asked to provide feedback and requirements
8
Provision and dissemination of
Observation Impact Statements
(OIS)
Lesson learnt:
Reactivity
Need to provide up-to-date results
Simple messages
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
• Definition of Observation Impact Statements
9
Provision and dissemination of
Observation Impact Statements
(OIS)
OISs are intended to be the primary
mechanism by which the OSEval-TT
will communicate findings of the
impact of observations on analysis
and forecast systems to observational
agencies
OISs are short reports that include a
standard set of graphics and tables
that can be easily understood by
stakeholders that are not experts in
data assimilation
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Evaluation studies
Perform impact studies of GOOS and ROOS on forecast and
reanalysis GODAE systems
OSE, OSSE and alternative methods
Provision and dissemination of Observation
Impact Statements (OIS) based on OSEval
evidence
Synthesis of impacts studies
Promote observation impact studies as core operational
activities for all ocean forecast centers
feedbacks
Requirements
support
Get the most
out of the obs
to improve the
GOV syst.
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Routine
Monitoring of the
GOOS
Delayed-mode
assessments of the
GOOS
Design and
evaluation of new
and future
observing system
components
• GOV OSEval-TT solution was to set up Near-Real-Time Observing System Experiments;
• It consists in running in parallel to the nominal operational forecast system and a second system where a single observation component of the GOOS will be with-held; and
• Implemented and tested by UKMet group in 2011#, under tested by Mercator-océan and Bluelink.
# Lea, D. J., M. J. Martin, P. R. Oke 2013: Demonstrating the complementarity of observations in an operational ocean forecasting system. Quarterly Journal of the Royal Meteorological Society, in press.
Synthesis of impacts studies
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Feb 2011
Mar 2011
Apr 2011
May 2011
Jun 2011
July 2011
XBT
TAO
Jason-2
All altims
SST
Argo
• paper in press by Lea et al. (2013; QJRMS)
• Methodology needs to be refined to take into
account: time influence of a data type,
dependence to a system (model/assimilation)
• Difficulties: relatively costly to maintain two
operational systems in parallel
Routine monitoring of the GOOS:
NRT OSEs
Temperature differences
for July 2011
Impact of ARGO
30 m
100 m
Data are with-held for one month -2°C +2°C
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Routine monitoring of the GOOS:
Alternative methods
Forecast sensitivity method using the NRL HYCOM system: these diagnostics show the importance of all observation types
(blue positive impact … improves the forecast; red negative impact … degrades the forecast)
Argo Temperature Altimetry XBT
Total Data Impact
For Jul-Aug 2012
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Routine monitoring of the GOOS:
Alternative methods Degree of freedom of Signal (DFS) tested by several groups
(Bluelink, TOPAZ, Mercator-Océan, DUACS NRT System, …)
Background and analysis error estimates (e.g., Bluelink)
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Routine monitoring of the GOOS:
Alternative methods Degree of freedom of Signal (DFS) tested by several groups
(Bluelink, TOPAZ, Mercator-Océan, DUACS NRT System, …)
Background and analysis error estimates (e.g., Bluelink)
Variability Analysis error
Salinity error at 100 m
Impact of Argo array
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Routine monitoring of the GOOS:
Alternative methods Degree of freedom of Signal (DFS) tested by several groups
(Bluelink, TOPAZ, Mercator-Océan, DUACS NRT System, …)
Background and analysis error estimates (e.g., Bluelink)
Pre-Argo
(univariate) Salinity error at 100 m
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Routine monitoring of the GOOS:
Alternative methods Degree of freedom of Signal (DFS) tested by several groups
(Bluelink, TOPAZ, Mercator-Océan, DUACS NRT System, …)
Background and analysis error estimates (e.g., Bluelink)
Demonstrating the additional constrain on salinity of the Argo
Post-Argo
(univariate) Salinity error at 100 m
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Routine
Monitoring of the
GOOS
Delayed-mode
assessments of the
GOOS
Design and
evaluation of new
and future
observing system
components
Synthesis of results
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
• OSEs using the Bluelink system with 0, 1, 2, and 3 altimeters;
• 1/10O-resolution - EnOI - Australian regional model
(showing 90th percentile of absolute differences between OSEs) (P. Oke)
The 1st altimeter has biggest impact – but the 2nd and 3rd altimeter provide
additional constraint 19
Impact of satellite altimetry June 2009
Independent XBT
3ALT
2ALT
1ALT
0ALT
NoDA
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
• OSEs using the REMO system
• 1/4o resolution - EnOI - Atlantic Basin
(showing RMSD with Argo profiles) (C. Tanajura)
Argo is the only data set to constrain salinity; and has the biggest impact on temperature
Impact of Argo, SST and altimetry
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Skill increases with more Argo floats
More floats may further improve skill
Impact of Argo • OSEs using the JMA-MRI MOVE-G
system – 1/3o-1o resolution – 3dVar
(showing the change in anomaly correlation: OSEArgo-OSEAll)
(Y. Fujii)
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Impact of TAO and Argo • OSEs using the JMA-MRI MOVE-G system – 1/3o-1o resolution – 3dVar
(showing the change in anomaly correlation: OSEArgo-OSEAll) (Y. Fujii)
Ch
an
ge
to
an
om
aly
co
rre
latio
n
Assimilation of Argo and TAO data improves forecasts on 1-
13 month times-scales
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
TAO data improves forecasts by ~0.4o in the Tropical Pacific
Impact of TAO, altimetry and Argo • OSEs using the ECMWF
system – ORSA4 (M. Balmaseda)
TAO data improves forecasts by ~0.4o in the Tropical Pacific
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Impact of loss of all data
• OSEs using the Canadian Global Ice Ocean Prediction System (GIOPS) - 1/4o resolution global model – Mercator’s SEEK/EnOI system (G. Smith)
(showing global RMS difference for temperature)
The model “forgets” about assimilated data over
several months.
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Impact of XBT • U Miami / NOAA (V. Kourafalou)
(showing impact on ocean heat content)
No airborne
profiles
1 deg XBT
sampling
0.5 deg
XBTs only
to 400 m
0.5 deg XBTs to
1000
m+altim+SST
High-res XBT data constrains features that
are not resolved by altimetry
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Conclusions from OSEval studies
• Satellite altimetry: regarded as the most critical data type for short-range mesoscale ocean forecasts – multiple satellite are needed
• Argo: regarded as the most critical data type for seasonal prediction – is the only data set to constrain salinity – evidence that more floats may return additional benefit;
• TAO: Improves seasonal forecasts (1-13 months);
• Data outages: Models “forgets” about assimilated data over several months
• XBT: significant local impacts – constrain features that are not resolved by altimetry
• Several studies highlight that different data types are complimentary
• Very few OSSE – lower priority than OSEs
• No Impact studies for SMOS and Aquarius
(See poster session for detailed results)
26
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Overall Summary
GOV OSEval-TT is an active community that is growing. Almost all the GODAE partners belongs to the TT;
New members give us the opportunity to revisit the impact studies(with new independent systems);
GOV OSEval-TT try to support observation agencies in demonstrating
the impact of observations on forecast and analysis products; Thanks to the different activities defined within the TT (NRT & DT
OSEs and alternative studies – DFS, …) and commitment from TT members we are able to cover large part of the demand from observational agencies;
Promote evaluation studies as a core activities of the GOV systems.
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Future
Continue to advocate to support observing system agencies;
Disseminate results of OSEval studies to observational agencies
Workshop: next one in Sept-Oct 2014
Assess impacts (or potential impacts) of biogeochemical observations; SAR altimetry; SWOT, Deep Argo …
Improve and develop new diagnostics to assess the observation impact depending of the applications (temporal and spatial scale of interest)
Improve and develop approaches to assess the contribution of one component of the observing system in the context of the whole observing system (relative weight of each component)
OSSE – coordinated activities … or not?
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Acknowledgements
Yosuke Fujii3, Emlyn M. Jones1, Vassiliki Kourafalou4, Ann Kristin Sperrevik5, Gregory C. Smith6, Daniel J. Lea7, Fiona Carse7, Baptiste Mourre8, Stephanie Guinehut2, Elisabeth Remy9, Clemente A. S. Tanajura10,11, Magdalena Alonso Balmaseda12, George R. Halliwell, Jr.13, Matthieu Le Hénaff4, Robert Atlas13, Andrew M. Moore11, Christopher A. Edwards11, Matthew J. Martin7, Alistair A. Stellar7, Alberto Alvarez8, Sandrine Mulet2, Isabelle Pujol2, Jim Cummings14
1CSIRO Marine and Atmospheric Research, Hobart, Tasmania, Australia 2CLS-Space Oceanography Division, Ramonville-Saint-Agne, France 3JMA/MRI, Tsukuba, Ibaraki, Japan 4University of Miami/RSMAS, Miami, FL USA 5Norwegian Meteorological Institute, Oslo, Norway 6Meteorological Research Division, Environment Canada, Quebec, Canada 7Met Office, Exeter, UK 8Centre for Maritime Research and Experimentation, La Spezia, Italy 9Mercator Ocean, Ramonville-Saint-Agne, France 10 Physics Institute, Federal University of Bahia (UFBA), Salvador, Brazil 11Department of Ocean Sciences, University of California, Santa Cruz, CA USA 12European Centre for Medium Range Weather Forecasts, Reading, UK. 13NOAA/AOML, 4301 Rickenbacker Causeway, Miami, FL, 33149, USA 14NRL, Monterey, CA, USA
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Impact of Coastal biogeochemical moorings
• EnOI “footprints” using MODIS satellite data (E. Jones)
Weather-band
(10-30d)
Intra-monthly
(30-90d)
Intra-seasonal
(90-180d)
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Impact of Coastal biogeochemical moorings
• EnOI “footprints” using MODIS satellite data (E. Jones)
A sparse array of moorings provides a
decent chance to monitor intra-sesonal
BGC variability
Weather-band
(10-30d)
Intra-monthly
(30-90d)
Intra-seasonal
(90-180d)
Isolated moorings over the GBR provide litte
constraint of short-time-scale BGC variability
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Routine monitoring of the GOOS:
Alternative methods Degree of freedom of Signal (DFS) tested by several groups
(Bluelink, TOPAZ, Mercator-Océan, DUACS NRT System, …)
Background and analysis error estimates (e.g., Bluelink)
Demonstrating the additional constrain on salinity of the Argo program
Pre-Argo
(multi-variate)
GODAE OceanView Symposium, Hilton Baltimore, 4-6 November 2013
Post-Argo
(multi-variate)
Routine monitoring of the GOOS:
Alternative methods Degree of freedom of Signal (DFS) tested by several groups
(Bluelink, TOPAZ, Mercator-Océan, DUACS NRT System, …)
Background and analysis error estimates (e.g., Bluelink)
Demonstrating the additional constrain on salinity of the Argo program