Rolf Sundberg: Statistical aspects on comparisons of climate simulation models and measured data (temperatures and temperature proxies)

Several people from mathematical statistics, SU, are now involved in research together with climate scientists from the Bert Bolin Centre for Climate Research. Vetenskapsrådet has supported work of mine and of Gudrun Brattström, and Katarina Fetisova will do her Ph.D. in the field, with Gudrun as principal supervisor.

The aim of my talk now is to give the audience an introduction to, and hopefully some feeling for, the statistical problems when different climate models, simulated during past periods of time, should be compared. The aim is to find the one that is in best agreement with actual data from the past. The models could for example differ in how they incorporate known variation in solar activity in the past. Data considered come from temperature measurements and from surrogates of such measurements, called proxies, for example tree rings data. Proxies go further back in time but need to be calibrated. Statistical methodology will be based on a joint model for simulated data, real temperature data, and proxy data.