Development of a 3-D Plasmapause Model With a Back-Propagation Neural Network

Abstract

Several empirical models have been previously developed to study the characteristics of the global plasmasphere. A three-dimensional solar wind-driven global dynamic plasmapause model was developed in this study using a back-propagation neural network based on multisatellite measurements. Our database contains 37,859 plasmapause crossing events from 4 January 1995 to 31 December 2015 covering all 24 magnetic local times (MLTs) from -66° to 79° magnetic latitudes. This model is parameterized by solar wind speed Vsw, interplanetary magnetic field Bz, SYM-H, and AE indices with smooth MLTs and latitude dependence. As the first 3-D empirical model, this model corresponds to the true plasmapause structure and is useful for observing space weather and other applications especially for predicting the shape of the plasmapause by forecasting the input parameters. Moreover, the proposed model is not only highly sensitive to these parameters, but also to their changes over days, seasons, and years; this means that the diurnal, seasonal, and annual variations of the parameters can be captured by the model. Therefore, this model can provide a more accurate plasmapause position compared with previous models and can also be used as a basic reference to develop other models such as a global plasmaspheric density model.