CMUG Climate Modelling User Group

1
CMUGClimate Modelling User Group

• Roger Saunders
• Met Office Hadley Centre

2
Overview and Meeting Aims

• Some key points from climate modelling
  perspective
• Meeting aims
• Inputs and outputs
• Wider perspective

3
CMUG is here to facilitate
Sea-ice
Sea-level
Sea surface temperature
Ocean Colour
Glaciers and ice caps
Land Cover
Fire disturbance
Cloud properties
Ozone
Aerosols
Greenhouse Gases
Climate Modellers
Reanalyses
4
CMUG folks here
Met Office Hadley Centre HadGEM, FOAM, HadISST
X
Paul Van Der Linden
Roger Saunders
Mark Ringer
ECMWF IFS, ERA, MACC
MPI-Meteorology ECHAM, JSBACH
MétéoFrance Arpege, MOCAGE, CNRM-CM, Mercator
X
Dick Dee
Thierry Phulpin
Alex Loew
Serge Planton
David Tan
Silvia Kloster
Stefan Kinne
Iryna Khlystova
5
Issues for climate modelling

• Higher resolution (horiz, vertical, time)
• Regional climate prediction (e.g. UKCP)
• More physical processes
• Seasonal to decadal prediction
• Use of reanalyses for climate
• Seamless prediction - weather prediction to
  climate change using same model
• Metrics developed to evaluate models CCI
  datasets can help here
• The way we use observational data is evolving

6
Climate monitoring and attribution
Different groups can produce defensible, but
statistically inconsistent estimates of trends.
Need for better error characterisation
7
Error characterisation of CDRs

• An estimate of the errors for each CDR produced
  is essential for use in climate applications
• The types of errors recently defined by GCOS
• Accuracy The rms difference between the single
  or averaged values of a variable and the truth.
• Stability The extent to which accuracy of a time
  average remains constant over a longer time
  period (e.g., annual average relative to decadal
  average).
• The importance of specifying each depends on the
  application
• Errors should be specified on a FOV basis.
  Aggregated error estimates are not sufficient
• Single sensor products are simpler than merged
  products
• Error correlations are also important to document

8
Use of observations evolving..
Observation simulator

• Forward modelling of measured quantities
  (radiances, skin SST, radar reflectivities)
  rather than high-level products (profile
  retrievals, bulk SST, cloud properties)
• Ensures more direct comparison of equivalent
  model variable with observations
• This was the key for use of ISCCP clouds

9
Multi-model analysis using satellite simulators
HadGEM1 (MO)
MMF 4km (CSU)
CloudSat
MMF 1km (CSU)
LMDZ (CNRS)
dBZgt-25
(Bodas-Salcedo et al., submitted to BAMS)
10
Implications for requirements

• The new ECV datasets must have added value over
  existing ones and future proof for model
  evolutions
• Datasets should have global coverage and for some
  applications gt15 years
• Be clear about applications for specific dataset
  as this drives the required accuracy
• Climate trend monitoring high stability
  and accuracy
• Change detection high
  stability
• Evaluate processes in model high accuracy
• Model validation high
  stability and accuracy
• Assimilation high
  accuracy (and stability)
• Uncertainty estimates are as important as product
  itself for all applications. Correlation of
  errors in space/time also important

11
Validation of SST
Coverage of buoys
Buoy validation of ARC SST
But what about ocean colour?
12
Meeting Aims

• Check ECV project URDs are consistent with the
  needs of Climate Research Groups and GCOS
  requirements, including source traceability
• Allow ECV teams to explain how their projects
  address the integrated perspective for
  consistency between the ECVs to avoid gaps
• Start review of product specifications
• Discuss how to deal with uncertainties in
  products
• Finalise the ECV projects data needs for ECMWF
  reanalysis data
• Start a discussion on ECV data set validation
• Maintain oversight of the position within the
  international framework in which CMUG/CCI is
  operating

13
URDs Common Issues

• CMUG report on CCI URDs D2.1
• Define period of TCDRs (1 month-30 years?)
• Clear specification of requirements for which
  application
• Some ECVs need clearer error specifications
• Merged vs single sensor products
• More interaction with climate modellers in some
  cases
• Consideration of model simulators where required
• Consistency between ECVs

14
Integrated view of ECVs

1. Through ensuring common input datasets are used
   for CDR creation and in some cases common
   pre-processing (e.g. geolocation, land/sea mask,
   cloud detection)
2. Through comparisons of CDRs for different ECVs
   (e.g. SST, sea-level, sea-ice and ocean colour)
3. Through comparisons of CDRs with model fields
   (e.g. GHG and Ozone CDRs and MACC model
   profiles/total column amounts) CMUG will be
   involved in development of some observation
   simulators. Pre-cursors of ECVs will be used for
   preparation.
4. Through studying teleconnections (e.g. El-Nino
   SST shows consistent impact on cloud fields,
   fires).
5. Through assimilation of CDRs and to assess impact
   on analyses and predictions (e.g. SST in
   ERA-Interim)

15
Outputs from meeting

• Meeting report of actions agreed by ECV projects
  including updates to URDs and Product Spec.
  docs
• Meeting report describing strategic position of
  the CMUG, within CCI, in the international arena
  
• Material to inform revision of CMUG reports
• Clarity on requests for ECMWF reanalysis data
• Clarity on early demonstration of products (if
  feasible) to modellers.

16
Related Activities

1. GCOS, GSICS (Jan/Feb 2011)
2. EUMETSAT CAF/CMSAF and SCOPE-CM
3. NOAA-NASA initiatives (e.g. JPL CMIP5)
4. WCRP Observation and Assimilation Panel (Apr 11)
5. Reanalyses (ERACLIM, JRA-55, EURO4M)
6. Coupled Model Intercomparison Project and
   follow-on activities (Exeter, June 11)
7. Inputs to IPCC AR-5/6 (interaction with authors)
8. EU IS-ENES, METAFOR,
9. EU GMES (MACC, MyOcean, Climate, .)

17

We dont want to leave our climate
research scientists like this!
But like this!
18
Any questions?Please visitwww.cci-cmug.org
cmug_at_metoffice.gov.uk
19
Proposed CMIP5 model runs
Proposed CMIP5 model runs
CCI datasets could start to be used in the
evaluation of these results
AR-5
20
Example of different errors
Stability 0.05K/decade
Accuracy 0.19K
SST
Bias 0.1K
Time
Buoy
Representativity and sampling