Mia Tian, Massachusetts Institute of Technology ; William Parker, The Aerospace Corporation; Richard Linares, Massachusetts Institute of Technology
https://doi.org/10.64861/RCQM7874
Keywords: satellite drag, conjunction assessment, space situational awareness, space weather, L1, LEO, DSCOVR, data fusion, space debris, thermosphere
Abstract:
The rapid growth in the number of satellites operating in Low Earth Orbit (LEO), combined with the planned deployment of large commercial constellations, has significantly increased the demand for accurate orbit prediction capabilities. Reliable trajectory forecasting is essential for conjunction assessment._x000D_
_x000D_
At altitudes below approximately 800 km, the dominant source of trajectory prediction error arises from thermospheric drag. Variations in solar extreme ultraviolet (EUV) irradiance and geomagnetic disturbances often initiated by coronal mass ejections (CMEs) inject energy into the upper atmosphere. These processes drive thermospheric expansion and density enhancements, significantly increasing drag forces experienced by spacecraft. Accurate, uncertainty-aware characterization of thermospheric environmental forcing on operationally relevant timescales remains a challenge for improving space situational awareness. Such characterization must provide sufficient lead time to enable operators to implement effective risk reduction maneuvers._x000D_
_x000D_
Response Estimation and Analysis for Correlated Trajectories (REACT) introduces a data-driven alternative that directly infers satellite drag response from linked observations without density modeling or knowledge of object properties. Conventional methods require propagating spacecraft through a forecasted thermospheric density field using satellite-specific properties–in practice, both are poorly estimated, rendering these methods inaccurate especially during space weather phenomena. REACT learns relationships between space weather observations and spacecraft dynamics to produce low-latency trajectory inference. By bypassing poorly observed intermediate steps, REACT provides a scalable and operationally responsive framework for trajectory prediction under dynamic thermospheric conditions._x000D_
_x000D_
This work initiates a forward prediction effort by using upstream space weather drivers to provide lead time. This work focuses on using Lagrange Point 1 (L1) data, which can offer hours of advance warning of geomagnetic forcing at Earth. This study ingests real-time solar wind and IMF measurements from DSCOVR and uses machine learning and statistical methods to estimate satellite decay rates. The trajectory forecasts are compared to historical catalog data to analyze accuracy and predictive power._x000D_
_x000D_
This study represents the first forward prediction application of REACT, and it demonstrates the value of integrating heterogeneous observations and in situ measurements within a unified predictive architecture. The resulting capability directly supports improved conjunction assessment and re-entry prediction in drag-dominated regimes._x000D_
Date of Conference: September 15-18, 2026
Track: Atmospherics/Space Weather