A new method for verifying deterministic predictions of meteorological scalar fields

In numerical weather prediction most error statistics with respect to forecasts of air pressure systems suffer from an intrinsic blending of “shape” errors with location errors. Separation of both effects is desirable. A method is presented which appears to be suited to meet this objective as far as short-term predictions are concerned, neglecting possible rotational effects. The scheme has not yet been in use operationally, but the results of some test cases are discussed in this paper. No reference is made to probability predictions nor to any economic purposes of evalution from the users' point of view, since the only aim is the scientific one of checking the behaviour of numerical models.