Thermospheric Density

Accurate prediction of thermospheric density is crucial for precise satellite orbit modeling and collision avoidance, particularly in low Earth orbit (LEO). Current research focuses on improving density models using machine learning techniques, such as neural networks (including neural ordinary differential equations and recurrent neural networks), to incorporate high-resolution solar imagery and improve upon existing empirical models like NRLMSIS 2.0. These advancements aim to reduce uncertainties in density predictions by leveraging the power of data-driven approaches and providing calibrated uncertainty estimates, leading to more reliable space situational awareness and improved satellite operations. The development of probabilistic models also allows for a better understanding of forecast uncertainty, enhancing the reliability of predictions.