Bayesian Classification of Geoeffective Solar Wind Structures

Abstract

“Space weather” refers to the condition of geospace, a plasma-filled region primarily consisting of the magnetosphere and the ionosphere. Space weather is controlled by the solar wind impinging on the outer boundary of the magnetosphere. Sporadic eruptions at the Sun such as coronal mass ejections and solar flares produce solar wind structures that can cause disturbances in the Earth's plasma environment, leading to adverse consequences on technological systems. Perhaps the most damaging of such “space weather effects” are severe geomagnetic storms characterized by intense disturbances that are long-lasting and global in geographic scale, encompassing both highlatitude and low-latitude regions around the Earth. Large geomagnetic storms are relatively infrequent occurrences but can seriously disrupt space-borne as well as ground-based susceptible systems including communications networks and electric power grids. Storms are often accompanied by increased populations of high-energy charged particles in geospace that can jeopardize satellites and astronauts. Thus, accurate and timely prediction of large storms is one of the most important end products of space weather research. The present prediction methods in operation, which are based on solar observations, are inaccurate because the trajectories and the magnetic fields of the ejecta from solar eruptions cannot be accurately predicted. This article describes the basics of a new approach to making real-time predictions of large geomagnetic storms. The approach is based on recognizing, in real-time solar wind data, quantifiable physical features that allow one to estimate the duration and geoeffectiveness of the solar wind that has yet to arrive at the detector. The results of an extensive test of the method using the archival WIND data from 1995–1999 indicate that high prediction accuracy (≥; 70–80 %) and moderately long warning times (several hours to more than ten hours) are achievable.