Automatic detection of the thermal electron density from the WHISPER experiment onboard CLUSTER‐II mission with neural networks
Abstract
The WHISPER (Waves of HIgh frequency and Sounder for Probing Electron density by Relaxation) instrument has been monitoring the bulk properties of the plasma environment around Earth for more than twenty years. Onboard the 3‐D Earth magnetospheric CLUSTER‐II mission, this experiment delivers active and natural electric field spectra, in a frequency interval ranging respectively from 3.5 to 82 kHz, and from 2 to 80 kHz. The thermal electron density, a key parameter of scientific interest and major driver for the calibration of particles instrument, is derived from spectra.
Until recently, the extraction of the thermal electron density required a manual intervention.
To automate this process, self‐learning algorithms based on Multilayer Neural Networks have been implemented. The evaluation of the thermal electron density from WHISPER spectra depends on the plasma region encountered by the spacecraft. First, a fully‐connected neural network has been implemented to predict the plasma region, using only the active spectra measured by the WHISPER instrument. Secondly, a specific neural network has been implemented to predict the thermal electron density for each plasma region. The model reaches up to 98% prediction accuracy for some plasma regimes. Two thermal electron density prediction models were trained, a first one to process data from the free solar wind and magnetosheath regions, and a second one for the plasmasphere region. The prediction accuracy can reach up to 95% in the free solar wind and magnetosheath regimes, and 75% in the plasmasphere.
Origin : Files produced by the author(s)