Estimating Errors in Trajectory Forecasts Using Ensemble Predictions

Abstract

The usefulness of the European Centre for Medium-Range Weather Forecasts ensemble predictions to predict the reliability of trajectory forecasts is investigated. About 3000 ensembles of trajectories are computed with a three-dimensional trajectory algorithm, and predictors from these ensembles are compared with the error of the trajectory forecasts. This error is defined as the distance from the trajectory forecast to the trajectory based on analyses and is determined both for the operational forecast?based trajectory (FCT) and for the ?middle member? of the ensemble (MMT). The evaluated predictors are the spread of the trajectory ensemble and the deviation of the FCT from the ensemble. The results show a small but significant positive correlation between the trajectory errors and their predictors, particularly between the deviation predictor and the error in the MMT. The probability distributions of the errors in FCT and MMT are also compared. Both small errors and very large errors occur more frequently for FCT than for MMT. For run times less than 48 h, the mean error is smallest for the FCT, particularly if FCT strongly deviates from the ensemble, whereas for longer run times, the ?risk- and success-avoiding strategy? of the ensemble renders the MMT more accurate.