Alternative measures of potential predictability applied to ensemble simulations of seasonal land-surface climate

摘要

The potential predictability (PP) of seasonal climate at the land surface is of enormous human import, and therefore merits close investigation. The PP of a seasonal land-surface variable may be defined as the upper bound in mean forecast to be expected when the seasonal state of the oceans is known precisely (as when SST's are prescribed), but when the initial conditions are known imprecisely. The PP of a seasonal land-surface variable is related to its degree of insensitivity to the choice of initial conditions when the ocean boundary conditions are invariant. There are various ways of measuring this initial-condition insensitivity, two of which are utilized here. The PP of a seasonal mean variable can be estimated from an ensemble of repeated simulations of a mul period in which the ocean boundary conditions are the same, but in which the initial conditions of the model's land/atmosphere system are different. It should be noted that a truly accurate determination of PP following this approach requires use of a perfect model. Estimates of PP made with current generation models there fore must be regarded as quite imprecise.