Pol Mesalles-Ripoll, SpaceNav; Roman Rositani, SpaceNav; Matt Duncan, SpaceNav
Keywords: atmospheric model, space weather, LEO, space situational awareness, predictive ephemeris
Abstract:
As the catalog of space objects in low-Earth orbit (LEO) grows and we witness an increasing number of conjunction events, owner/operators find themselves in need of generating accurate orbit predictions for the purposes of space situational awareness.
Although the 18th and 19th Space Defense Squadrons of the U.S. Space Force already generate ballisitc trajectories for all trackable objects for operational conjunction assessment, not all operators have access to these ephemerides, and in many cases they might require more frequent predictions that they can use for planning, tasking, and scheduling. Additionally, for all maneuverable satellites, it is of the utmost importance that they share predicted ephemerides with the rest of the space community to communicate their intent to maneuver and allow other operators to plan accordingly.
Many satellite operators already leverage on-board GNSS receivers to obtain definitive solutions and produce their own orbit predictions, but there is a lack of resources on which atmospheric models are best suited for the application of generating real-time predicts in LEO.
In this paper, we evaluate the prediction accuracy of three different modern empirical atmospheric models: NRLMSISE-00, NRLMSIS 2.0, and JB2008.
The Naval Research Laboratory (NRL) mass spectrometer and incoherent scatter radar (MSIS) extended to the exosphere (E) from the year 2000, or NRLMSISE-00 for short, is one of the most widely used atmospheric models when propagating LEO orbits. It takes in the day of year, altitude, latitude, and longitude as main inputs, as well as the F10.7 solar flux index (previous day and 81-day average) and geomagnetic activity (daily Ap and 3-hour ap index for 0, 3, 6, 9, 1233, and 3657 hours before current time).
The NRLMSIS 2.0 model is a reformulation upgrade of NRLMSISE-00 first introduced in 2020 that takes in essentially the same inputs as the previous version. When developing the model, the NRL focused on altitudes below 200 km, but the incorporation of new measurements and slight changes to the thermosphere modeling have significant effects on state propagation.
The Jacchia-Bowman atmosphere, last updated in 2008 (JB2008), is an empirical thermospheric density model that uses Jacchia’s diffusion equations and new solar indices from Space Environment Technologies (SET): F10, S10 (EUV), M10 (MUV), and Y10 (X-ray). Additionally, it takes into account geomagnetic storm effects using the Dst index. In addition to space weather, it takes in as inputs the modified Julian date, the geocentric latitude, longitude, and height of the spacecraft, and the right ascension and declination of the Sun.
In addition to the atmosphere models, we also investigate the effects of two different sources of space weather data: daily predictions retrieved from the NOAA Space Weather Prediction Center and 3-hourly predictions from SET (known as JBHSGI indices), which are the same inputs that the U.S. Space Force uses for their own High-Accuracy Atmospheric Satellite Drag Model (HASDM).
For this study, we have selected two satellites that operate in different orbit regimes: the first one is in a 400×390 km, 65° very low altitude orbit, while the second one is flying in a Sun-synchronous orbit at 650 km altitude, but has a higher area-to-mass ratio.
To evaluate the performance of each atmospheric input, we cannot simply compare generated ephemerides with the same initial conditions and different atmosphere models; parameters such as the drag coefficient must also be estimated with the same model before generating new predictions. Thus, we have set up an end-to-end flight dynamics system where we process new GNSS NavSOL tracking data for each satellite multiple times per day through SpaceNav’s in-house orbit determination (OD) service to generate definitive ephemerides. With the final state vector and estimated spacecraft parameters, we then generate a new 7-day predictive ephemeris using the same model from the OD process. In both cases, the state propagation is handled by our in-house software, which includes support for the NRLMSISE-00, NRLMSIS 2.0, and JB2008 atmospheric models, and always uses the most recent space weather data from the selected source.
By aggregating the results over several months and overlapping predicted ephemerides on top of the previously generated definitives, we can determine which model produces statistically the best results. For each atmosphere model and each date and time over the span of the study, we have a definitive state (position and velocity with covariance) and multiple predicted states (the latest result, as well as the predicts from 17 days before). This large dataset allows us to compare the position and velocity differences in the radialin-trackcross-track (RIC) frame for each model at different prediction lengths. Additionally, we can evaluate that the definitive ephemerides obtained by each model are self-consistent (within the uncertainty of each other) and that consecutive predicted ephemerides for a given model are also self-consistent.
Date of Conference: September 19-22, 2023
Track: Atmospherics/Space Weather