Application of quasi-random ensemble calculations for determination of climate model optimal parameters

By analyzing a randomly generated set of runs, each 2000 years in length, the author has considered the uncertainty in 12 mixing and transport parameters. Constructing a quantitative measure for the model error made it possible to address both the inverse problem of estimation of model parameters and the direct problem of model predictions. The results represent an attempt at tuning a three-dimensional climate model by a strictly defined procedure which, nevertheless, considers the whole of the appropriate parameter space. The modeling approach is thus to match model outputs to observations while model inputs (parameters) are initially only weakly constrained.