SYSTEMATIC BIASES IN THE OBSERVED DISTRIBUTION OF KUIPER BELT OBJECT ORBITS

The orbital distribution of Kuiper Belt objects (KBOs) provides important tests of solar system evolution models. However, our understanding of this orbital distribution can be affected by many observational biases. An important but difficult to quantify bias results from tracking selection effects; KBOs are recovered or lost depending on assumptions made about their orbital elements when fitting the initial (short) observational arc. Quantitatively studying the effects and significance of this bias is generally difficult, because only the objects where the assumptions were correct are recovered and thus available to study "the problem," and because different observers use different assumptions and methods. We have used a sample of 38 KBOs that were discovered and tracked, bias-free, as part of the Canada-France Ecliptic Plane Survey to evaluate the potential for losing objects based on the two most common orbit and ephemeris prediction sources: the Minor Planet Center (MPC) and the Bernstein and Khushalani (BK) orbit fitting code. In both cases, we use early discovery and recovery astrometric measurements of the objects to generate ephemeris predictions that we then compare to later positional measurements; objects that have large differences between the predicted and actual positions would be unlikely to be recovered and are thus considered "lost." We find systematic differences in the orbit distributions which would result from using the two orbit-fitting procedures. In our sample, the MPC-derived orbit solutions lost slightly fewer objects (five out of 38) due to large ephemeris errors at one year recovery, but the objects which were lost belonged to more "unusual" orbits such as scattering disk objects or objects with semimajor axes interior to the 3: 2 resonance. Using the BK code, more objects (seven out of 38) would have been lost due to ephemeris errors, but the lost objects came from a range of orbital regions, primarily the classical belt region. We also compare the accuracy of orbits calculated from one year arcs against orbits calculated from multiple years of observations and find that two-opposition orbits without additional observations acquired at least two months from opposition are unreliable for dynamical modeling.