Empirical methods have become crucial for producing climate scenarios for e.g. impact assessment. Improved methods for statistical downscaling of climate scenarios, as well as statistical refinement of output from regional models, have been developed during the later years. Various groups have developed different techniques and tested out different strategies, predictors & predictands, optimal domain sizes, etc. We now believe it could be fruitful for representatives from different groups working with these problems to meet in order to 1) summarize what we have learned, 2) point out central problems which still need to be solved and eventually suggest how to deal with them, and 3) to discuss the possibility of future cooperation, e. g. in a "Statistical Model Intercomparison Project" (SMIP) where we suggest standards for comparisons of our methods.


We plan a 2-day workshop with 3 introductory presentations addressing the workshop themes. Other participants are invited to give short (10-15 min) prepared presentations, but most of all we want an open discussion where everyone contribute with their experience and expertise. To make this setting efficient, we believe that the number of participants should be restricted to about 20. The results from the workshop should be a report summarizing status, recommendations and problems concerning statistical downscaling, and possibly also plans for future cooperation.

Possible themes for discussion (in arbitrary order)

  1. Internal consistence for statistically downscaled scenarios with respect to covariation between different variables, and/or to autocorrelation in space and/or time. Advantages/ disadvantages for different methods.
  2. Extremes: Are some indicators better/ more robust than others? Methods?
  3. Domain size in statistical downscaling: How sensitive are the results to domain size? Is there an optimal choice? Is it possible to give general recommendations?
  4. Predictands and predictors: Can we agree on recommendations concerning predictors for local temperature and precipitation? What about local wind, snow cover/ water equivalent and other variables?
  5. Methods: Can we give general recommendations here, or is this totally dependent on the problem?
  6. Uncertainty associated with statistical downscaling.
  7. Aapplicability of statistical downscaling methods.
  8. "Statistical Model Intercomparison Project": Could we agree on some basic principles/ standards for comparisons of our models, e.g. in a common project (SMIP)?