introduction to cmug assessments, sst and plans for phase 2 roger saunders 4 th integration meeting
TRANSCRIPT
Introduction to CMUG assessments, SST and plans for phase 2
Roger Saunders
www.cci-cmug.org
4th Integration Meeting
CMUG Assessments of CCI CDRs
Why does CMUG assess CCI datasets?
• Provide an independent view of the datasets and associated uncertainties
• Study consistency between ECVs
• Demonstrate applications for climate modelling to accelerate use by the climate/reanalysis communities
CMUG assessments in phase 1
Methodology used
for assessment of ECVs
Assessment of precursors(see CMUG D3.1 report series)
Assessment of CCI CDRs(see CMUG D3.1 report series)
Climate Model (single, ensemble)
O3((IASI), Land Cover (GlobCover), SSH (AVISO), Cloud (ISCCP), Fire (GFEDv3)
O3, Land Cover, SM, SSH, Cloud
Re-analyses SST (HadISST), O3(ERA) O3, SM, Aerosols, GHG, SSH
Precursor datasets OC, SSH, SST, O3
Independent satellite or in situ measurements
SST (ARC) SST, O3, OC
Related observations (surface and TOA fluxes, temperature, water vapour)
Fire (GFEDv3 ) SM
Assimilation GlobColor OC
Data issues found and fixed
• Data gaps– Day: western Pacific and off western coast of India (only ATSR-1
[after failure of 3.7 μm] and ATSR-2), – Night: south-western Atlantic and first 7 months of 1993.
• Abrupt change in the number of retrievals during night at ~2oS and ~8oN.
• Fill values for depth SST and uncertainties (huge numbers for 1993, 2000 and 2002 [AATSR]).
• Days without data, e.g. 29th February for the leap years.
Assessment of CCI SST
• Main validation against drifting buoys• Comparison with precursor ARC dataset
Retrieval methods– CCI: optimal estimation for ATSR-2 and AATSR, but ATSR-1 as ARC.– ARC: coefficients from radiative transfer simulations.
• Different spatial resolution– CCI: 0.05 deg– ARC: 0.1 deg
• Same cloud screening method– minor changes due to updated versions of radiative transfer and NWP
models.
Bias: ATSRs-drifting buoys
Larger bias in the tropics during day (water vapour ?) and in the mid-latitudes and over Indian and western Pacific oceans during night (systematic errors ?).
CCI ARC
night
day
Bias: Hovmoller diagrams
• AATSR performs better than ATSR-1/2.• Oscillation with a warmer bias during the boreal winter months in mid-latitudes.• Surprising difference between CCI and ARC for ATSR-1, given the same
retrieval method.
CCI ARC
night
day
3-way error analysis
• Use TMI SST for validation of ATSR-2 in the zone [-40, 40]oN for the first time.• Practically, ATSR-2 and AATSR have the same performance with standard
deviation of error 0.18 K for ARC and 0.23 K for CCI. • Drifting buoys showing an improvement of their random error with time, while
also AMSR-E has better performance than TMI.
Method which provides the random error given that the three datasets (buoys, ATSRs and MWs [TMI and AMSR-E]) are uncorrelated.
Uncertainty assessment
ARC’s uncertainty assignment is in general better than CCI’s uncertainty, although not for all uncertainty values or validation criteria.
CCI ARC
night
day
No CCI uncertainty < 0.1 K.
Conclusions on assessment of SST CCI
• 2ch biases in SST significantly higher than for ARC
• 3ch bias in SST slightly higher than ARC
• Uncertainties of CCI product suggest they are reasonable but less matchups for nightime cases than for ARC
• Feedback suggests AVHRR SST dataset is an improvement over the pathfinder SST dataset. This is good news to extend the time series back before 1995.
Cross-ECV consistency
SST SL Cl Sice OC Aero GHG LV Fire Ozone Glaci IC SM
SST x x X X x x
Sea level x x x
Clouds x x X x x X x X
Sea ice X x x X x x
Ocean col X x x x
Aerosol x X X x X x
GHG x x x X
Landcover x x x x x
Fire x x x X x x
Ozone x x X
Glaciers x X x
Ice Sheets X
SoilM x x
Strong Weaker
CMUG phase 2 - Core ProposalCross-ECV dataset assessments• Cross-Assessment of marine ECVs (SST, OC, SSH, SI) (Met Office)
• Integrated assessment of the aerosol, GHG, and ozone datasets (ECMWF)
• Integrated exploitation of CCI terrestrial ECVs (LC, Fire, SM) (MPI-M/IPSL)
• Cross-Assessment of ECVs from sea-ice with atmospheric ECVs (MPI-M/DLR)
• Cross-Assessment of Aerosols, Cloud and Radiation CCI ECVs (DLR)
• Cross assessment of clouds, radn, aerosol, GHG, s moisture, SST (SMHI/MF)
Exploiting CCI products in CMIP like experiments• Assessing CCI datasets as boundary conditions in CMIP5-like atmosphere
simulations (MF/IPSL)
Adaption of community climate evaluation tools
• Benchmarking models with ESA CCI data in the era of CMIP6 (DLR/MPI-M)
• Development of community climate dataset evaluation tools (ECMWF)
Interface to climate services
Discussion After Coffee
Phase 1 Outcomes:• What additional value / problems compared to the
CMUG results were identified in climate research groups of the different ECV teams?
• What were major limiting factors in data usage identified by the CRG's?
Phase 2 and CCI-2 plans• How will phase-2 address limitations of phase-1
data identified by CMUG/CRG's ?
• What should be the new variables/products in a potential CCI-2? (e.g. Snow, Lakes, Salinity, Albedo, …….)