Solar wind observations show that geomagnetic storms are mainly driven by interplanetary coronal mass ejections (ICMEs) and corotating or stream interaction regions (C/SIRs). We present a binary classifier that assigns one of these drivers to 7,546 storms between 1930 and 2015 using ground‐based geomagnetic field observations only. The input data consists of the long‐term stable Hourly Magnetospheric Currents index alongside the corresponding midlatitude geomagnetic observatory time series. This data set provides comprehensive information on the global storm time magnetic disturbance field, particularly its spatial variability, over eight solar cycles. For the first time, we use this information statistically with regard to an automated storm driver identification. Our supervised classification model significantly outperforms unskilled baseline models (78% accuracy with 26[19]% misidentified interplanetary coronal mass ejections [corotating or stream interaction regions]) and delivers plausible driver occurrences with regard to storm intensity and solar cycle phase. Our results can readily be used to advance related studies fundamental to space weather research, for example, studies connecting galactic cosmic ray modulation and geomagnetic disturbances. They are fully reproducible by means of the underlying open‐source software (Pick, 2019, http://doi.org/10.5880/GFZ.2.3.2019.003)
Plain Language Summary
The Earth's magnetic field reaches out into space where it constantly interacts with the solar wind, a stream of charged particles from the Sun. Geomagnetic storms occur when discontinuities in the solar wind disturb the geomagnetic field, possibly causing failures of, for example, electricity transmission and sat...

A Statistical Classifier for Historical Geomagnetic Storm Drivers Derived Solely From Ground‐Based Magnetic Field Measurements

10.1029/2019EA000726