Neural network-based calibration of Swarm Langmuir Probe ion densities

    The European Space Agency's Swarm constellation consists of three spacecraft (A, B, and C). Each of the satellites is equipped with a Langmuir probe (LP), which measures ion densities and temperatures. The LP processing assumes that the plasma consists exclusively of oxygen ions, which leads to the nighttime overestimation of plasma densities due to non-negligible influence of light ions that is not accounted for in the LP processing. Each of the Swarm satellites also provides electron density measurements by the Faceplate (FP), which is part of the Thermal Ion Imager (TII) suite. The FP densities do not depend on assumptions of the plasma composition. In this study, we use the FP densities as a reference, in order to calibrate the LP observations. We model the ratio of FP to LP data using neural networks. We create three models, for each of the satellites, and are able to produce, even from sparse observations, correction factors for Swarm LP densities. The proposed correction exhibits significant variations based on local time, season, altitude, and solar activity, consistent with the presence of light ions due to the downward ambipolar diffusion from plasmasphere. The developed model resolves the nighttime overestimation by Swarm-LP. The corrected LP data are in excellent agreement with COSMIC radio occultation observations, and can be used for numerous applications including empirical modeling of the topside ionosphere.