Gauging model performance:

1
Gauging model performance

• a)     Should a validation standard be
  established?
• b)    If so, what measurements might be part of
  the standard?
• c)     Can we suggest inter-comparison
  protocols?
• - Bromwich/Proshutinsky (CG Center Bay)

2
1)     Should a validation standard be
established?

• Yes, this is an important issue and has to be
  based on (a) observational model quality and (b)
  model errors. This is a specific task for
  post-MIP studies. MIPs have to determine model
  errors and then to recommend validation standards
  (parameters, grids, spatial and temporal time
  scales).
• Observers and modelers have to be involved in the
  establishment of these standards
• Benchmarks have to be (a) integrative, (b) small
  in number, (c) well constrained observationally

3
2. If so, what measurements might be part of the
standard?

• We have already discussed these parameters for
  all disciplines yesterday and today (atmosphere,
  ice, ocean, terrestrial, etc.)
• Each discipline has to identify its most
  important components, What are the major
  integrated measures/processes which have to be
  reproduced correctly (seasonal cycle, AO
  variability, sea ice, etc.) for all disciplines.

4
2b. If so, what measurements might be part of the
standard?

• Atmosphere (how to reduce matrix of benchmarks?)
• 1. Large-scale integrative parameters/processes
• Mean state and variability patterns
• Storm track positions
• Precipitation (as an example of a crucial
  parameter)
• 2. Reproduce several important parameters and
  processes
• Validate annual energy balance at surface
  (fluxes, clouds, stratification, interactions
  with sea ice and ocean).
• 3. To validate this use drifting stations, remote
  technology.
• 4. How to validate models against these
  benchmarks?
• 5. What are the observational errors?

5
Ocean

• Ocean and Ice
• Integrative processes sea ice drift, extent, and
  seasonal cycle, distribution of sea ice
  thickness
• mean ocean parameters (vertical stratification
  in the European and Canada basins), heat, salt
  and mass transports through najor straits, Fram
  Strait, Barents Sea Opening, Canadian Straits and
  Bering Strait,
• Fresh water surface layer, cold halocline, warm
  Atlantic layer, deep cold waters
• Correct fluxes via straits and shelf-basin
  exchanges

6
Terrestrial

• Integrating parameters/processes
• Albedo
• Annual river runoff
• Soil temperatures/permafrost distribution
• Vegetation distribution should be linked to
  dynamical vegetation model
• 2. Individual processes/parameters
• Active layer depth variability
• Snow depth and density (snow water equivalent)
• 3. Data availability
• - Satellites for albedo, vegetation, snow
  parameters
• Representative station data snow, precipitation,
  soil data
• Processed gridded data (maps based on section
  serveys)

7
Can we suggest inter-comparison protocols?

• Yes, a set of model intercomparison protocols
  already exists in different MIPs.
• For AOMIP we have a common grid, and a list of
  integrated parameters to be inter-compared. Could
  be based on expert estimates.