Reconstructing Substorms via Historical Data Mining: Is It Really Feasible?

The evolution of the low‐latitude magnetosphere over the substorm cycle is reconstructed based on a new high‐resolution 3D representation of the magnetic field and nearest‐neighbor data mining. The study covers radial distances 2.5–25RE and employs a record‐large pool of spacecraft data taken during 1995–2019. The magnetospheric state is quantified by four indices, representing the ground geomagnetic activity and its temporal trends in the entire ±90° range of geomagnetic latitude: the SuperMAG SMR, the midlatitude positive bay MPB, the auroral SML, and the polar cap PC index. The developed technique has been tested on specific substorm events, with the results presented in the form of 5‐min cadence diagrams and animations of the magnetic field line configurations and electric current distributions. In all the analyzed events, the initial intensification and radial expansion of the inner tail current is accompanied by a gradual stretching of the magnetic field, followed by its sudden collapse, dramatic depletion of the current beyond R∼12RE, and a large‐scale dipolarization of the field around the time of MPB peak, after which the system recovers and tends to its pre‐substorm state. Plain Language Summary The dynamical structure of the Earth's magnetosphere during geomagnetic substorms is reconstructed, based on (a) a large multi‐year database of satellite data taken during the last quarter century, (b) a pool of concurrent ground geomagnetic activity indices, covering full range of latitudes, (c) a new magnetic field model with enhanced spatial resolution, and (d) an advanced “nearest‐neighbor” data mining approach. Based on a synthesis of the above methods and data, we explore the ability of our approach to extract maximum information from past observations and reproduce the principal phases of magnetospheric substorms in terms of time sequences of the magnetic field and electric current diagrams, from the beginning to active and recovery phase of the disturbance. Key Points A new high‐resolution B‐field representation combined with dynamical data mining reveals magnetosphere behavior on the substorm‐time scale Full cycle of magnetosphere evolution is reconstructed based on 25‐year archive of satellite data and a set of ground‐based activity indices Initial growth of the distant magnetotail current and its subsequent sudden collapse during the substorm onset are consistently reproduced