A method is presented for the verification of numerical forecasts of the position and intensity of high and low pressure systems and the identification of systematic biases which may occur in the output from a numerical analysis-assimilation-prediction scheme (hereafter referred to as output). The method is applied primarily to the output from the five-layer primitive equations Spectral Model which was first used operationally by the National Meteorological Analysis Centre, Melbourne, Australia in 1976, but a comparison between the Spectral Model and the Sub-synoptic Advection Model (SAM) output is also given. The analysis of the spectral model output is made in the region 10°E to 179°E for low pressure systems and between 20°W and 179°E for high pressure systems; the base data set covers the period July 1978 to February 1979, The analysis confirms and quantifies systematic biases that have become apparent to forecasters through routine operational use. The systematic biases include the tendency displayed in the model output to under-forecast the intensity of high and low pressure systems, to underestimate the eastward movement of high and low pressure systems, and to move highs too far equatorward. It is found that the biases could be adjusted for by simple regression equations that correlated changes predicted in the model output to those that are actually observed. Using these equations a significant reduction can be achieved in the root mean square error of forecasts of both central pressure and movement of the systems. For the Sub-synoptic Advection Model output (based in part on the Spectral Model) the results indicate that for the limited number of cases studied within the Australian region the magnitude of both the overall and systematic biases is reduced and that regression equations can be derived that will further reduce the biases.
Pescod et al. (Mon,) studied this question.