Probabilistic Drag-Based Ensemble Model (DBEM) Evaluation for Heliospheric Propagation of CMEs

Abstract

The Drag-based Model (DBM) is a 2D analytical model for heliospheric propagation of Coronal Mass Ejections (CMEs) in ecliptic plane predicting the CME arrival time and speed at Earth or any other given target in the solar system. It is based on the equation of motion and depends on initial CME parameters, background solar wind speed, \(w\) and the drag parameter \(\gamma \). A very short computational time of DBM (< 0.01 s) allowed us to develop the Drag-Based Ensemble Model (DBEM) that takes into account the variability of model input parameters by making an ensemble of n different input parameters to calculate the distribution and significance of the DBM results. Thus the DBEM is able to calculate the most likely CME arrival times and speeds, quantify the prediction uncertainties and determine the confidence intervals. A new DBEMv3 version is described in detail and evaluated for the first time determining the DBEMv3 performance and errors by using various CME–ICME lists and it is compared with previous DBEM versions, ICME being a short-hand for interplanetary CME. The analysis to find the optimal drag parameter \(\gamma \) and ambient solar wind speed \(w\) showed that somewhat higher values (\(\gamma \approx 0.3 \times 10^{-7}\) km−1, \(w \approx \) 425 km s−1) for both of these DBEM input parameters should be used for the evaluation than the previously employed ones. Based on the evaluation performed for 146 CME–ICME pairs, the DBEMv3 performance with mean error (ME) of −11.3 h, mean absolute error (MAE) of 17.3 h was obtained. There is a clear bias towards the negative prediction errors where the fast CMEs are predicted to arrive too early, probably due to the model physical limitations and input errors (e.g. CME launch speed). This can be partially reduced by using larger values for \(\gamma \) resulting in smaller prediction errors (\(\mathrm{ME} =-3.9\) h, MAE = 14.5 h) but at the cost of larger prediction errors for single fast CMEs as well as larger CME arrival speed prediction errors. DBEMv3 showed also slight improvement in the performance for all calculated output parameters compared to the previous DBEM versions.