This project will identify and minimise the tendency - or forcing - errors in four different atmospheric general circulation models (GCMs). In this way new model versions will be developed which are improved relative to the basic versions, in the sense that the total forcing errors are reduced without introducing compensations between multiple errors. The improved models will be developed and tested with special attention to simulation of regional climate over Europe and to seasonal prediction of climate. The four models are two state of the art atmospheric climate models and two simpler GCMs.

Different techniques will be used to calculate the forcing errors, but the basic method, to be tested in all four GCMs, consists of a simple four dimensional data assimilation technique called nudging. In nudging a given model is constantly relaxed towards data which varies in time. In the present application these data will be the ECMWF Re-Analyses (ERA) and the magnitude of the relaxation will with different constraints be used as a measure of the models’ forcing errors.

The work will have an iterative character where improvement of the GCMs will be guided by the temporal and spatial distribution of the forcing errors and by the characteristics of the systematic errors seen in climate runs.

The improvement of the GCMs will follow two lines:

1. Improvement of physical parameterization. This improvement will be validated with special attention to simulation of regional climate over Europe and elsewhere since such errors severely degrades the performance of regional climate models which receive boundary conditions from the GCMs.
2. Developments of empirical parameterization of the forcing errors. Here the focus will be on testing the seasonal prediction capabilities of GCMs which include an empirical parameterization or correction of their forcing errors.