Key points are not available for this paper at this time.
Abstract In this study, the thermospheric mass density (TMD) features observed by the CHAllenging Minisatellite Payload between 2002 and 2010 were extracted using deep learning (DL) technology; the TMD features were then mapped and modeled with the Interplanetary environment information (IEI), solar radiation, and geomagnetic indices. The DL model was used to simulate the TMD features during Day of Year (DOY) 222–241 in 2014, a period that experienced complex solar‐terrestrial environmental variations. We explore the TMD features under different solar‐terrestrial environmental conditions and discuss the effects of various inputs by comparing the DL simulation results with satellite observations from Gravity Recovery and Climate Experiment‐A and Swarm‐A, as well as the simulation results from Jacchia‐Bowman 2008, Naval Research Laboratory Mass Spectrometer Incoherent Scatter radar model 2.1, and Drag Temperature Model 2013. These results show that the DL model can better capture the TMD features after adding IEI. Part of these TMD features, including the high‐latitude TMD enhancement during the space hurricane event (DOY 232, 2014) and global TMD variations under complex solar‐terrestrial environmental disturbances (DOY 222–225, 2014), cannot be well described by the geomagnetic indices. The DL model indicates that the east‐west component of the interplanetary magnetic field (IMF By) has a great impact on TMD variations, and its modulation is different from the typical energy injection process during storms. Our results emphasize the crucial influence of IEI on TMD under both geomagnetic disturbances and quiet conditions.
Li et al. (Wed,) studied this question.