Assessment and Categorization of TLE Orbit Errors for the US SSN Catalogue

The Space Debris Office at ESA predicts conjunction events based on Two-Line Element (TLE) data obtained from the US Space Surveillance Network. Currently two ESA missions, the Low-Earth orbiting satellites ERS-2 and Envisat, are covered. For all conjunction events that passed a so-called smart sieve filtering the related collision risk is assessed and provided in a bulletin that is distributed by email daily. In case a high-risk conjunction event is forecast external tracking data of the chaser are acquired. Orbit determination using these data gives improved state and covariance information of the chaser. A subsequent re-assessment of the collision risk allows to decide on the necessity of collision avoidance maneuvers and to support the planning of necessary maneuvers. At ESA’s Space Debris Office the central tools for analyzing conjunction events are the collision risk assessment software CRASS and the orbit determination software ODIN. The risk assessment faces the problem that no covariance information is available for the TLE data set. CRASS copes with this issue by introducing pre-defined look-up tables for the initial covariance that are sorted by eccentricity, perigee height, and inclination. Through ODIN the covariance information is obtained from comparing states derived directly from the TLE data with states resulting from an orbit determination using pseudo-observations derived from TLE data. The obtained covariance information reflects the limitations of the TLE (SGP4) orbit model combined with the limitations of ODIN in terms of orbit determination and propagation accuracy. Until now the CRASS look-up table contains only a limited number of orbit classes. Recently, a new command-line version of ODIN has been developed, allowing repetitive, fully automated analyses. Thus, the application of the covariance estimation procedure to the entire TLE catalogue becomes feasible. We address the orbit determination and propagation quality of ODIN by comparing orbits determined from precise radar tracking with external high-precision orbits obtained from laser-tracking and Doppler ranging, and by comparing propagated states to theses high-precision orbits. For a current catalogue we assess the TLE orbit errors in alongtrack, cross-track, and out-of-plane coordinates (i.e. as function of eccentricity, inclination and perigee height). This analysis provides a more realistic look-up table for the collision risk assessment with CRASS. Insights into the applicability of the TLE theory to certain classes of orbits will be helpful in particular for the selection of data product formats for the European Space Situational Awareness system that is under study. Finally the presented approach may be the basis for comparisons of snapshots of the TLE catalogue of past epochs.