Evaluating Coupled Climate Models: Roles of Global Observational
  Networks and Local High-Density Experiments.

   Progress in Coupled Climate Models' fidelity requires essentially 
 two tasks: validation of the model results and improving the 
 representation of physical processes in the models. These two tasks
 create very different challenges for both the modeling and the
 observational communities. Validation requires global, long term
 data acquisition and detailed knowledge of statistics. In
 contrast, improving the representation of physical processes
 often requires short, high density observations, and
 physical intuition. Arguably, these differences in goals and 
 methods can explain why the needs of the modeling community
 are met by observations only accidentally. For any meaningful
 progress on both tasks, there is a need for these two 
 communities to communicate and collaborate more effectively. 
 To achieve this goal, our primary focus is to bring
 junior people from both communities together in this forum.
 The topics for discussions will include the observational data
 needs of the climate modeling community, what observations provide,
 what is missing, and the limitations of models and observations.   

   As a specific and relevant example, one can focus on ENSO. 
 In spite of decades of research, coupled climate models fail to reproduce
 the observed mean, seasonal cycle, and interannual variability of the
 tropical Pacific climate. As a basis for the discussions we hope that
 the senior scientists (Kathie Kelly and Tony Busalacchi) will provide
 an outline of the current state of research in this (and other areas) 
 and where they expect that the biggest improvements in the future 
 will come from.

   Using ENSO as an example, the problems can roughly be tied 
 to two things: large uncertainties
 in the observed climate (e.g. rainfall, mixed layer depths) and limited
 understanding of the key physical processes (e.g. convection, oceanic
 vertical mixing). Given the limited resources how should the community
 proceed? For example, is it realistic to expect that we can bypass the
 shortcomings of the mean state and still improve on the representation
 of the interannual variability? Shall we give up on the direct obser-
 vations of rainfall or vertical mixing and just infer them with sen-
 sitivity studies or an adjoint approach? A larger question is: Do we
 understand ENSO and therefore only need to fix model details, or is
 the systematic failure of the models an indication that something fun-
 damental is missing?

   Again, just to restress, we use ENSO as a sample problem,
 and any other "trouble" areas are OK to discuss and expose. 

   The main goal of the workshop is to facilitate interaction between junior
 scientists from different institutions. We hope that after
 the workshop all the attendees have a common understanding of the problems
 facing progress in climate research, reevaluate their own
 approaches, and share the dark secrets of their own subdiscipline without
 fear from senior scientists.

    Markus and Gokhan