A More Extensive Investigation of the Use of Ensemble Forecasts for Dispersion Model Evaluation

Abstract

An ensemble forecast is used as input to a Lagrangian particle dispersion model to study the effect that analysis errors in the numerical weather prediction assimilation cycle have on dispersion modeling. The wind and temperature fields from a control forecast are first compared with radiosonde measurements to quantify the quality of the forecast. It predicts, on average, a less stable atmosphere near the surface, whereas the average deviations in wind speed and wind direction are small. The dispersion model then simulates a tracer puff release from the European Tracer Experiment (ETEX), by using the ensemble forecast as input. The spread in the ensemble forecast is calculated by comparing the predicted puff concentrations for each ensemble member with those from the control forecast. It was found that ensemble members that were close to the control forecast when one statistical measure was calculated were not necessarily close when another was used. The ensemble spread is then quantified by comparison with measurements from ETEX. The ensemble spread is also compared with the spread resulting from the model formulation by comparing the model results with the results from 34 models participating in ETEX. The ensemble spread is found to be important after about 24 h after the release started, whereas the spread in performance attributable to the model formulation is important throughout the simulation period. Last, the possible use of ensemble forecasts in emergency management is discussed. At present it is too time consuming to use ensemble forecasts for emergency management purposes.