Major Charity Weeden, NORAD-USNORTHCOM/J55X
Keywords: NROC, Non-Resolved Object Characterization
Abstract:
At the 2005 AMOS conference, Kriging Optimized Interpolation (KOI) was presented as a tool to model satellite brightness as a function of phase angle and solar declination angle (J.M Okada and M.D. Hejduk). Since November 2005, this method has been used to support the tasking algorithm for all optical sensors in the Space Surveillance Network (SSN). The satellite brightness maps generated by the KOI program are compared to each sensors ability to detect an object as a function of the brightness of the background sky and angular rate of the object. This will determine if the sensor can technically detect an object based on an explicit calculation of the objects probability of detection. In addition, recent upgrades at Ground-Based Electro Optical Deep Space Surveillance Sites (GEODSS) sites have increased the amount and quality of brightness data collected and therefore available for analysis. This in turn has provided enough data to study the modeling process in more detail in order to obtain the most accurate brightness prediction of satellites. Analysis of two years of brightness data gathered from optical sensors and modeled via KOI solutions are outlined in this paper. By comparison, geo-stationary objects (GEO) were tracked less than non-GEO objects but had higher density tracking in phase angle due to artifices of scheduling. A statistically-significant fit to a deterministic model was possible less than half the time in both GEO and non-GEO tracks, showing that a stochastic model must often be used alone to produce brightness results, but such results are nonetheless serviceable. Within the Kriging solution, the exponential variogram model was the most frequently employed in both GEO and non-GEO tracks, indicating that monotonic brightness variation with both phase and solar declination angle is common and testifying to the suitability to the application of regionalized variable theory to this particular problem. Finally, the average nugget value, or prediction fidelity, was slightly better in the non-GEO tracks than in GEO, often centering around half of a visual magnitude. In describing these relationships between modeling variables and a successful brightness solution, better predicted brightness values for use in operational tasking can be produced.
Date of Conference: September 10-14, 2006
Track: Non-Resolved Object Characterization