Solar-terrestrial energy has a crucial influence on the composition and dynamics of the high-latitude thermosphere (Prölss, 2011;Richmond & Lu, 2000). The thermosphere is affected fundamentally through solar extreme ultraviolet (EUV) radiation and magnetospheric energy input. EUV energy input is the dominant driver in quiet time, but for high-latitude thermosphere during geomagnetic storms, magnetospheric energy can far exceed it (Knipp et al., 2004;Prölss, 2011). Furthermore, Joule heating and particle precipitation are the two main forms of magnetospheric energy, while during geomagnetic storms Joule heating can count for up to two-thirds of the energy being deposited into the thermosphere (Knipp et al., 2004). During the most active times, the mechanism of injection energy transfer from the solar wind to the coupled magnetosphere-ionosphere-thermosphere (CMIT) system is ultimately dissipated into the thermosphere through strong Joule heating (

The neutral mass density variation associated with the induced satellite drag force plays an important role in the Low Earth Orbit (LEO) spacecraft operations in the thermosphere, such as orbit maintenance, lifetime, and collision avoidance (Doornbos, 2012;Krauss et al., 2018;Zesta & Huang, 2016). Due to the impact of the solar-terrestrial energy input from the magnetosphere to the ionosphere-thermosphere system, the thermospheric density has intricate spatial and temporal variations. Consequently, predicting and simulating the neutral mass density changes and the related drag feedbacks to the spacecrafts are still a grand challenge for space weather research.Since the 1950s, several empirical thermospheric density models have been applied for orbit determination and prediction of LEO spacecraft, such as Jacchia model and Mass Spectrometer Incoherent Scatter (MSIS) model (Jacchia, 1971;Hedin, 1987Hedin, , 1991. However, the density discrepancies between the observations and these