Item

Abstract

Earth’s magnetotail is a huge reservoir where the energy of the solar wind-magnetosphere interaction is accumulated and then suddenly released during substorms. It is also important as the largest space region available for in-situ observations of key processes in collisionless plasmas, such as magnetic reconnection. Yet until recently, the rarity and sparsity of observations made the first-principles theory and simulations of the magnetotail very difficult to verify with data, or even use observations to guide theoretical models. The situation has changed due to implementation of the machine learning methods, and in particular, mining multi-decade and multi-mission arrays of spaceborne magnetometer data. Here we show how mining 25+ years of such data allows one to reconstruct the magnetotail during substorms. The reconstruction turns out to be so accurate that it captures not only stretching and dipolarization of the tail during a substorm but it also pinpoints the location of the X- and O-lines associated with the substorm reconnection. It is seen from the comparison of their location with a 4-year long collection of ion diffusion regions provided by the Magnetospheric Multiscale mission. This empirical or “black-box” information can then be merged with global MHD simulations (“white-box models”) in the form of empirical resistivity maps whereby forming a “grey-box model”. It can be further advanced by adding the empirical information on the location and dynamics of thin ion-scale current sheets, key structures providing the energy accumulation in the tail, as well as their white-box simulations in kinetic particle-in-cell codes.