Figure 2. (Left) Observation of OneWeb-0210 obtained with the 0.6-meter telescope at Chungbuk National University Observatory, South Korea (courtesy of CLEOsat). The field of view is 72 × 72 arcminutes and the exposure time is two seconds. The image center is indicated by the blue crosshairs, while the red crosshairs show the forecast position of the satellite from a TLE. The red spot indicates the satellite’s true position, which is 15.1 arcminutes off from the forecast position. (Right) A reference image from the ESO digital sky survey[1] centered on the TLE forecast position, which allows the telescope pointing uncertainty to be corrected for. The red crosshairs are in the same location in both images.
However, obtaining a robust measurement of TLE accuracy requires more than a single observation. One such attempt was performed by the Pomenis team led by Harrison Krantz (University of Arizona). Over 560 observations were conducted in the summer and autumn of 2020. Their analysis compared the trail centroid (assuming constant angular velocity) with the predicted TLE position of the satellite. Specifically, they measured the full uncertainty in the satellite positions, which is the sum of three error vectors: TLE accuracy, telescope pointing, and the uncertainties from the orbital equations. They concluded the statistical uncertainty in TLE accuracy is ± 3 arcminutes. Their results also show that the distribution tail extends beyond 0.5 degree, as shown in Figure 3, which is equivalent to the angular diameter of the Moon and larger than many telescope fields of view. The work by the Pomenis team used both operator-provided TLEs directly and the supplemental TLEs derived from operator-provided ephemerides, and actually found no discernible difference in the accuracy between the two types of TLE.
