In Low Earth Orbit (LEO), atmospheric drag is the largest contributor to trajectory prediction error. The current thermospheric density model used by the Combined Space Operations Center (CSpOC) in operations is the High Accuracy Satellite Drag Model (HASDM). Since HASDM is not available for use outside of the US Government, satellite operators are left to determine what publicly available, open-source density model they should integrate into their internal operational software. This decision is nontrivial due to the number of available density models, each having variable performance dependent on several factors including space weather conditions and orbit altitude. To compound matters, the rapid rise of this solar cycle suggests that the predicted solar maximum between 2024-2027 could be higher than the previous solar maximum, thus causing larger perturbations due to drag from atmospheric density on LEO satellites. Given the evermore challenging nature of operations in LEO, it is imperative for satellite operators to update legacy density models to a state-of-the-art density model to provide improved trajectory predictions for collision risk assessment and vital day-to-day operational decisions. This paper outlines several operations-ready thermospheric density models, describing their performance, computation time, required operational space weather input parameters, and notes for implementation. We define an operations-ready density model as a model that is well-documented, has verified and quantified model performance, and provides publicly available model code for implementation on a user’s own system. Operations-ready models include the Drag Temperature Model (DTM), the Jacchia-Bowman 2008 (JB2008) model, the US Naval Research Laboratory Mass Spectrometer and Incoherent Scatter radar 2.0 (NRLMSIS 2.0) model, and the Thermosphere– Ionosphere–Electrodynamics General Circulation Model (TIE-GCM). US Government operational density models, HASDM and the Whole Atmosphere Model and Ionosphere Plasmasphere Electrodynamics (WAM-IPE) model, are included for comparison in the Analysis section. Models are evaluated against global HASDM density and local Gravity Recovery And Climate Experiment Follow-On (GRACE-FO) satellite accelerometer density data. A propagation analysis is also included in which model performance is compared during quiet and storm conditions and resulting LEO object trajectory prediction errors are quantified at various orbit altitudes. The analysis shows that any of the named operations-ready density models (DTM2020, JB2008, NRLMSIS 2.0, TIE-GCM) are a viable option for satellite operations. In addition to LEO satellite operators, the results from this paper are also informative for the transition of civilian space traffic coordination efforts out of CSpOC and into the Department of Commerce. @en

The satellite acceleration data from the CHAMP, GRACE, GOCE, and Swarm missions provide detailed information on the thermosphere density over the last two decades. Recent work on reducing errors in modelling the spacecraft geometry has greatly reduced scale differences between the thermosphere data sets from these missions. However, residual inconsistencies between the data sets and between data and models are still present. To a large extent, these differences originate in the modelling of the gas-surface interactions (GSI), which is part of the satellite aerodynamic modelling used in the acceleration to density data processing. Physics-based GSI models require in-situ atmospheric composition and temperature data that are not measured by any of the above-mentioned satellites and, as a consequence, rely on thermosphere models for these inputs. To reduce the dependence on existing thermosphere models, we choose a GSI model with a constant energy accommodation coefficient per mission, which we optimize exploiting particular attitude manoeuvres and wind analyses to increase the self-consistency of the multi-mission thermosphere mass density data sets. We compare our results with those based on variable energy accommodation obtained by different studies and semi-empirical models to show the principal differences. The presented comparisons provide novel opportunity to quantify the discrepancies between current GSI models. Among the presented data, density variations with variable accommodation are within ±10%, and peaks can reach up to 15% at the poles. The largest differences occur during low solar activity periods. In addition, we utilize a series of attitude manoeuvres performed in May 2014 by the Swarm A and C satellites, which are flying in close proximity, to evaluate the residual inconsistency of the density observations as a function of the energy accommodation coefficient. Our analysis demonstrates that an energy accommodation coefficient of 0.85 maximizes the consistency of the Swarm density observations during the attitude manoeuvres. Using such coefficient, for Swarm A and Swarm C, the new density would be lower in magnitude with a 4-5% difference. In recent studies, similar energy accommodation coefficients were retrieved for the CHAMP and GOCE missions by investigating thermospheric winds. These new values for the energy accommodation coefficient provide a higher consistency among different missions and models. A comparison of neutral densities between current thermosphere models and observations indicates that semi-empirical models such as NRLMSISE-00 and DTM-2013 significantly overestimate the density, and that an overall higher consistency between the observations from the different missions can be achieved with the presented assumptions. The new densities from this work provide consistencies of 4.13% and 3.65% between the minimum and maximum mean ratios among the selected missions with NRLMSISE-00 and DTM-2013, respectively. A comparison with the WACCM-X general circulation model is also performed. Similar to the other models, WACCM-X seems to provide higher estimates of mass density especially under high and moderate solar activities. This work has the objective to guide density data users over the multiple data sets and highlight the remaining uncertainties associated with different GSI models.

@en