摘要：The k-nearest-neighbor technique is used to mine a multimission magnetometer database for a subset of data points from time intervals that are similar to the storm state of the magnetosphere for a particular moment in time. These subsets of data are then used to fit an empirical magnetic field model. Performing this for each snapshot in time reconstructs the dynamic evolution of the magnetic and electric current density distributions during storms. However, because weaker storms occur more frequently than stronger storms, the reconstructions are biased toward them. We demonstrate that distance weighting the nearest-neighbors mitigates this issue while allowing a sufficient amount of data to be included in the fitting procedure to limit overfitting. Using this technique, we reconstruct the distribution of the magnetic field and electric currents and their evolution for two storms, the intense 17–19 March 2015 “Saint Patrick's Day” storm and a moderate storm occurring on 13–15 July 2013, from which the pressure distributions can be computed assuming isotropy and by integrating the steady-state force-balance equation. As the main phase of a storm progresses in time, the westward ring current density and pressure increases in the inner magnetosphere particularly on the nightside, becoming more symmetric as the recovery phase progresses. We validate the empirical pressure by comparing it to the observed pressures from the Van Allen Probes mission by summing over particle fluxes from all available energy channels and species.
