For the modelling subgroup of the TISO retreat:

o Modelling efforts often take one or more of the following paths:

  1. Writing new model, new code/science.
  2. Adding new code/science to existing models.
  3. Rewriting old code for existing or new models.
  4. Explicitly coupling 2 or more existing but separate models (MIMD).
  5. Merging existing models into a single code (SIMD).

  With examples of all of these processes occurring at hao, this
  is clearly an era of "model mergers". As such, model development
  should proceed with emphases on collaborative efforts between
  disciplines (e.g., atmospheric regimes) in science, software engineering, 
  and computer science, e.g.:

  1. Use of software frameworks and layering within which specialty
     (discipline specific) code can be developed
  2. Reusable code in different contexts (e.g., numerical utilities, libs)
  3. Common or flexible metadata data formats (input and output)
  4. Use of coding and language standards and conventions, 
     both local and national/international
  5. Flexible or common data and gridding structures
  6. Portability over different platforms and architectures
     (incluing shared and distributed memory system architectures)
  7. Semi-independent parameterizations that may be implemented
     in different model frameworks
  8. Inter-organizational collaborations

  HAO should pay particular attention to NCAR proposals to NSF's ITR
  program, and NCAR's multi-agency proposal to NASA for an "Earth System 
  Modeling Framework", see Staff Notes Monthly at 
  http://www.ucar.edu/communications/staffnotes/0010/modeling.html.
  See also, "NCAR's Strategic Plan for High Performance Simulation".