Statistical Corrections to Dynamical Model Predictions

Abstract

A statistical correction scheme based on standard 1east-squares linear regression substantially improves 24 h mean sea level pressure predictions for the Australian region. Mean-square error is reduced by 20% and S1 skill is improved by 3 points. The correction to the model prediction at a grid point is a weighted average of estimates from several regression equations. The predictors in these equations are grid point values from the base analysis and the 24 h prognosis. The regression equations were developed for the 254 km grid Australian Region Primitive Equations (ARPE) model, but were also successful in reducing the error of the 150 km grid Fine Mesh Model (FINEST).