A methodology for numerical prognosis evaluation using objective local weather retrieval

A general method of evaluating numerical model prognoses in terms of local weather forecasts using the perfect prognosis approach is described. The main advantages of the method are that clean impact studies are feasible and only a short prognosis archive period is required. Information is given concerning which prognosis fields need most improvement to benefit local forecasts. Using a novel skill score to relate the objectively derived weather forecasts to the operational ones, the methodology is applied, as an example, to the operational Australian region primitive equation low-level prognoses' impact on Sydney, January daily maximum temperature forecasts. The prognosis fields are shown to require improvement particularly in the low-level temperature fields before they can provide operationally useful forecasts of Sydney, January maximum temperatures.