Updating Measures of CME Arrival Time Errors

Coronal mass ejections (CMEs) drive space weather effects at Earth and through the rest of the heliosphere. Predicting their arrival is a major part of space weather forecasting. In 2013, the Community Coordinated Modeling Center started collecting predictions from the community, developing an Arrival Time Scoreboard (ATSB). Riley et al. (2018) analyzed the first 5 years of the ATSB, finding a bias of a few hours and uncertainty of order 15 hr. These metrics have been routinely quoted since 2018, but have not been updated despite the availability of continued predictions. We revise analysis of the ATSB using a sample 3.5 times the size of that in the original study. We find generally the same overall metrics, a bias of -2.5 hr, mean absolute error of 13.2 hr, and standard deviation of 17.4 hr, with only a slight improvement comparing between the previously-used and new sets. The most well-established, frequently-submitted model results tend to outperform those from seldomly-contributed models. These "best" models show a slight improvement over the 11 year span, with more scatter between the models during early times and a convergence toward the same error metrics in recent years. We find little evidence of any correlations between the arrival time errors and any other properties. The one noticeable exception is a tendency for late predictions for short transit times and vice versa. We propose that any model-driven systematic errors may be washed out by the uncertainties in CME reconstructions or in characterization of the background solar wind, and suggest that improving these may be the key to better predictions.