Alex Bush, OKAPI:Orbits GmbH; Adrian Diez, OKAPI:Orbits GmbH; Oscar Rodriguez Fernandez, OKAPI:Orbits GmbH; Sebastian Hesselbach, DLR, German Space Situational Awareness Centre
https://doi.org/10.64861/IKAS4677
Keywords: Orbit Determination, Space Weather, Uncertainty, Covariance Realism, VLEO, LEO, UKF, Kalman Filter
Abstract:
The increasing deployment of satellites into Low Earth Orbit (LEO) and Very Low Earth Orbit (VLEO), combined with growing operational demands for conjunction assessment, constellation management, and re-entry prediction, has made realistic orbit uncertainty modelling a critical capability. Conventional propagators either omit atmospheric drag uncertainty or rely on static, ad-hoc process noise matrices that cannot represent the structured, space-weather-driven character of thermospheric density errors. This limitation causes filter divergence and unreliable covariance estimates under disturbed conditions, with direct consequences for missed conjunction alerts, degraded station-keeping decisions, and unreliable re-entry forecasts. This paper presents OPUS (the Operational Parametric Uncertainty System), a parametric uncertainty modelling and calibration system embedded within the NEPTUNE numerical propagator developed alongside the German Aerospace Centre (DLR). OPUS is proposed as a physically-structured, dynamic alternative to manual process noise tuning. The parametric model decomposes total drag uncertainty into a superposition of physical components, each encoding positional, altitude, and external-variable (including space weather) dependencies through spherical-harmonic and PCHIP sub-models driven by pre-processed proxies. The framework is input-agnostic, any normalised space-weather or operational time series, including F10.7, geomagnetic storm indices, and forecast products, can be registered as a driver without structural changes to the model. Uncertainty is injected via a consider-parameter formulation intended to capture the coherent covariance growth driven by persistent thermospheric model errors, with a magnitude that varies with orbital geometry and space weather state rather than being held fixed across all regimes. An ephemeris-based calibration pipeline fits the model coefficients by optimising Mahalanobis distance consistency against a χ² reference, using a replay mechanism that makes the search tractable without repeated full propagation. Results are preliminary and demonstrate the calibration methodology on a single object.
Date of Conference: September 15-18, 2026
Track: Atmospherics/Space Weather