Abstract. We report the first systematic survey of cosmic ray precursors of geomagnetic storms. Our data set comprises the 14 "major" geomagnetic storms (peak Kp > 8-) identified by Gosling et al. [ 1990] together with 25 large storms (peak Kp > 7-) observed from 1992 through 1998. After eliminating events for which the muon detector network had poor coverage of the sunward interplanetary magnetic field (IMF) direction, we determined that 15 of the remaining 22 events (68%) had identifiable cosmic ray precursors with typical lead times ranging from 6 to 9 hours prior to the storm sudde.n commencement (SSC). Of the 15 precursors, 10 were of the "loss cone" (LC) type which is characterized by an intensity deficit confined to a narrow pitch angle region around the sunward IMF direction. Cosmic rays in the loss cone presumably originate in the cosmic-ray-depleted region downstream of the approaching shock. The remaining five precursors were of the "enhanced variance" (EV) type which is characterized by intensity increases or decreases that do not systematically align with the IMF direction. The incidence of precursors increases with storm size; for instance, 89% of storms with peak Kp greater than or equal to 8.0 had precursors. Our results show that the muon detector network can be a useful tool in space weather forecasting. However, new detector(s) installed to fill major gaps in the present network are urgently required for better understanding the nature of precursors and for reliable space weather forecasting. In this paper we examine cosmic ray precursors of large geomagnetic storms by analyzing high-energy cosmic ray intensities observed by ground-level muon detectors. As far as we know, this is the first systematic survey of cosmic ray precursors in which event selection is determined by the occurrence of a geomagnetic storm rather than by a Forbush decrease or other cosmic ray effect. Similar analysis using the high-latitude neutron monitor network will be reported hours prior to the shock arrival, assuming a shock speed of 600 km/s. This also implies that the precursors might be observed much earlier by muon detectors than by neutron monitors, as the muon detector responds to higher-energy (> 50 GeV) cosmic rays for which •.// is expected to be much longer. In this respect, the muon observations at higher-energy regions may supply useful information for forecasting geomagnetic storms. This is a primary motivation of the present analysis. It will be demonstrated that the muon observation network could be a very good tool for space weather forecasting if it has a good coverage of a full range of pitch angles.
Data Analysis MethodWe analyze the cosmic ray precursors of geomagnetic storms recorded by a network of ground-level muon detectors at "Nagoya" (Japan), "Hobart" (Australia) and "Mawson-PC" (Antarctica) (see Table 1). Each of these detectors is multidirectional, allowing us to simultaneously record the intensities in various directions of viewing. The total number of directional telescopes used in this...