A methodology of retrieving volume emission rate from 630.0 nm airglow emission intensity by combing the technologies of Abel inversion and deep learning

This study is a follow-on research of "The comparison and validation of photochemical models for atomic oxygen ion retrieval from ground-based observations of O(1D) 630.0 nm airglow near Irkutsk" by Duann et al. (pending), which revealed that the 630.0 nm emission inversion model derived atomic oxygen ion density ([O+]) in mid-latitudes F-region has similar features to both FORMOSAT-3/COSMIC and Irkutsk digisonde DPS-4 detected electron density (Ne) profiles. To further validate the applicability of the inversion model at low latitudes and equatorial regions with global satellite observations, the FORMOSAT-2/ISUAL observed 630.0 nm airglow emission intensity data are selected for this purpose. The 2009 solar minimum period is examined here to reduce the influence of the solar activity variation to the airglow emission intensity. This study developed a method to optimize the retrieval of satellite-based imager captured 630.0 nm intensity (in unit of Rayleigh) into volume emission rate (VER) by using the technology of deep learning. More specifically, the training was setup with 3 hidden layers but switched each layer with a different number of neurons, and the targets generated from the nets are the coefficient and the thickness for building a Chapman distribution. By converging the topside of each ISUAL observed profile to zero with deep learning Chapman distribution, the Abel inversion can be applied to determine volume emission rate profiles. In this study, a large number of nets generated from the training process are tested, and the nets with best performance with model results and real-time data are selected and compared. With the combination of the Abel inversion and the deep learning algorithm, the radio occultation inversion methodology has potential for optimizing the process of converting airglow emission intensity from Rayleigh into VER, and improves the capacity of analyzing ionospheric observations as well.