Vishal Ray, Kayhan Space; Eric Sutton, University of Colorado / SWx TREC; Jeffrey Thayer, University of Colorado; Siamak Hesar, Kayhan Space Corp; Derek Strobel, Kayhan Space
Keywords: Atmospheric density, HASDM, drag, calibration
Abstract:
With multiple satellite mega-constellations, such as Spire, SpaceX’s Starlink, OneWeb and Amazon’s Kuiper being planned in the LEO regime over the next decade, space traffic in the near-Earth space domain will be reaching unprecedented levels of congestion. Many of these planned constellations, such as Starlink, utilize automatic collision avoidance techniques for orbit safety and maintenance of their satellites. But they are still reliant on orbit data of all other satellites and debris in the neighborhood of their satellites from the 18th Space Squadron. Therefore, accurate prediction of the orbital states, and subsequently, thermospheric density, is a problem of growing significance that requires immediate attention. The effect of thermospheric density perturbations becomes especially evident in the very low earth orbit (VLEO) regime where small uncertainties in the density predictions can have dramatic impacts on operations as shown by the February 2022 Starlink loss event.
The current operational density model for orbit predictions and conjunction analyses used by the Department of Defense (DOD) is the Air Force High Accuracy Satellite Drag Model (HASDM). HASDM is considered the gold-standard of operational density modeling as it provides a very accurate picture of the current average state of the thermosphere by assimilating real-time tracking data from around 90 calibration objects and dynamically correcting the baseline JBH09 model (a slight variant of Jacchia Bowman 2008 or JB2008 model). But HASDM densities are only available in real-time for government users. For non-government users, there are several non-assimilating density models, such as the Mass Spectrometer Incoherent Scatter Radar (MSIS) series of models and JB2008, used in operations with no clear consensus on which model performs the best in what scenario. In general, these models are typically biased, either high or low, with HASDM containing the least biases due to real-time data assimilation. During quiescent conditions, these biases can be somewhat corrected during orbit determination resulting in acceptable orbit fits and short-term orbit predictions. But during extreme space weather conditions, the models fail to capture perturbations over short time-scales and local spatial-scales leading to poor orbit fits and possible catalog loss. As the provision of civil SSA services transition from DOD to DOC as per the Space Policy Directive – 3 (SPD-3), a commercial atmospheric density model with a performance comparable to HASDM or better will be essential for the smooth operation of civil SSA.
An accurate thermosphere forecasting system requires a model rooted in full physics of the coupling between the upper and lower atmosphere and the Sun-Earth system. Such a model would be better suited to account for local-scale and higher frequency perturbations that semi-empirical models fail to capture. Since physics-based models can be computationally quite complex, machine-learning methods can be utilized to make them operationally feasible. In order to be consistent with the current average state of the thermosphere, a physics-based model needs to be continuously calibrated with real-time data ingested through a data-assimilation methodology. The planned mega-constellations discussed in the preceding paragraph can serve as signals of opportunity for such a system with their extensive spatial and temporal coverage of the thermosphere. Additionally, planned science missions such as the Geospace Dynamics Constellation (GDC) will carry high-fidelity GPS receivers that can be used for high accuracy density retrievals along the orbit.
One of the first steps towards such a data-assimilative physics-based model of the thermosphere is accurate retrieval of near real-time densities. There are a few types of density retrieval methods utilizing different types of tracking information, with varying accuracy and complexity. These can essentially be classified under two groups – 1) processing tracking data in an orbit determination scheme and 2) using the tracking information directly to calculate densities. The former class of methods can be placed under the broad category of “accelerometry”, i.e., estimating non-conservative accelerations from tracking data in an orbit determination scheme that can further be used to calculate atmospheric densities or directly estimating densities from the data. The second class of methods can be used to calculate energy dissipation or accelerations by using the tracking data as it is, and densities can be inferred from the obtained quantities. In this paper, we explore both classes of methods by utilizing Precision Orbit Determination (POD) as measurements – 1. POD accelerometry and 2. Energy Dissipation Rate (EDR) calculation using POD ephemeris – and compare their sensitivities to various possible errors and tuning parameters. The relative contributions of the various error sources – truncation of gravitational field, nongravitational forces, and attitude – to the estimated densities are studied. We also discuss possible ways to mitigate these error sources. Since the POD noise levels and their arc-lengths can vary depending on how they have been obtained (e.g., kinematic, reduced-dynamic or dynamic) and the duty cycle of the GPS receiver, the accuracy of the estimated densities is analyzed for different POD noise levels and measurement cadence. The results of this study can be used to predict the uncertainties around estimated densities when processing satellite tracking data across various altitudes.
The ultimate goal of this work is to create a long-term, continuously updated, neutral density dataset that can be ingested in a “commercial data-driven atmospheric density modeling framework”. Our preceding analyses will inform the relative advantages and ease of implementation of the two density retrieval methods. We finally compare the retrieved neutral density estimates against HASDM densities using data from the Spire constellation and limited data from the Starlink constellation. Our results demonstrate that the outputs from density models (semi-empirical or physics-based) are in much better agreement with HASDM densities when calibrated by satellite tracking data.
Date of Conference: September 19-22, 2023
Track: Atmospherics/Space Weather