Research Article – invited by the Editors-in-Chief

Verification of space weather forecasting at the Regional Warning Center in Belgium

Royal Observatory of Belgium (ROB) and Solar-Terrestrial Centre of Excellence (STCE), Ringlaan 3, 1180 Brussels, Belgium

^* Corresponding author: andy.devos@oma.be

Received: 21 March 2014
Accepted: 11 September 2014

Abstract

The Solar Influences Data analysis Center (SIDC) in Brussels at the Royal Observatory of Belgium (ROB) has been providing daily space weather forecasts for more than a decade. A verification analysis was applied to evaluate the performance of the SIDC forecasts of fundamental space weather parameters such as the F10.7 radio flux, solar flare activity, and local geomagnetic index.

Strengths and weaknesses are determined compared to common numerical models. Descriptive model statistics, common verification measures, error analysis and conditional plots related to forecasts and observations are presented. The verification analysis methods have been designed such that future improvements and additions can easily be included, for example with new forecasting models.

The SIDC forecast (together with the persistence model) achieves the best performance for forecasting F10.7 on day 1, but has potential for improvement for a larger lead time mainly by applying estimates from the persistence and corrected recurrence models. The persistence model is superior for the forecast of flares, though corrected recurrence models are slightly better in foreseeing M- and X-class flares and the SIDC forecast estimates B- and C-class flares very well. The SIDC forecast scores better than all models in forecasting the local K-index. It best reproduces observations in the range of K = 2–4, but underestimates larger K values. The SIDC forecast provides a distribution that best matches the observations of the K-index. The analysis presented here demonstrates the influence of solar activity on the confidence level of the forecasts, as well as the hinted influence of the forecaster on duty due to the subjective nature of forecasting. The output aids to identify the strong and weak points of the SIDC forecast as well as those of the models considered. Though the presented analysis needs further extension, it already illustrates the opportunity to regularly reevaluate space weather forecasts and to stimulate ideas for improvement and increase the reliability of space weather forecasting.