Geomagnetic storm forecasting service StormFocus: 5 years online

TL;DR

This paper reviews the five-year performance of the StormFocus geomagnetic storm forecasting service, which predicts storm strength using real-time solar wind data, highlighting its accuracy and data discrepancies.

Contribution

It provides an evaluation of StormFocus's real-time prediction accuracy over five years and analyzes the impact of data discrepancies on forecast reliability.

Findings

87% storm prediction accuracy in real-time

97% accuracy with final OMNI data in reanalysis

Data discrepancies are main source of prediction errors

Abstract

Forecasting geomagnetic storms is highly important for many space weather applications. In this study we review performance of the geomagnetic storm forecasting service StormFocus during 2011--2016. The service was implemented in 2011 at SpaceWeather.Ru and predicts the expected strength of geomagnetic storms as measured by $Dst$ index several hours ahead. The forecast is based on L1 solar wind and IMF measurements and is updated every hour. The solar maximum of cycle 24 is weak, so most of the statistics are on rather moderate storms. We verify quality of selection criteria, as well as reliability of real-time input data in comparison with the final values, available in archives. In real-time operation 87% of storms were correctly predicted while the reanalysis running on final OMNI data predicts successfully 97% of storms. Thus the main reasons for prediction errors are discrepancies between real-time and final data (Dst, solar wind and IMF) due to processing errors, specifics of datasets.