Improved Atmospheric Density Estimation and Characterization of Uncertainty

Precise knowledge of the density in the upper atmosphere is a vital component of the orbit determination process for low Earth orbit, as inaccuracies in the estimation of atmospheric drag are the primary source of uncertainty for satellites in low Earth orbit. The need for a more accurate knowledge of the density of the upper atmosphere has led to the development of atmospheric density derived from precision satellite orbits. This method, using the Precise Orbit Ephemerides (POE) for a satellite, requires refinement and validation before it can be used on a larger scale. Additionally, the uncertainty of this method is not well documented. To improve these atmospheric density models, the POE densities are calculated and compared to the accelerometer derived densities for a majority of the lifetime of both the Challenging Minisatellite Payload (CHAMP) and Gravity Recovery and Climate Experiment (GRACE) satellites to provide a more robust understanding of the effectiveness of these models. Additionally, the framework has been set so that future satellite missions can easily be ingested and analyzed without a substantial amount of work. In a few locations in the accelerometer derived densities, there are gaps that must be filled. By using a separate accelerometer density method, the first, more reliable accelerometer method can be patched in these locations to allow for a much more robust method. This combined density allows for a more effective evaluation of the POE densities. To further improve the estimates of atmospheric density, the four Atmospheric Neutral Density Experiment (ANDE) satellites are considered. These spherical satellites provide a much simpler analysis of the atmospheric drag than the much more complicated geometry of the established CHAMP and GRACE satellites. To further improve the estimates for the ANDE satellites, a series of methods to more accurately model the drag coefficients for these satellites are studied and applied to the orbit determination process. In addition to the ANDE satellites, the CHAMP and GRACE satellites drag coefficients were updated to include a higher fidelity drag coefficient and projected area model. Using these drag coefficients, the atmospheric densities are estimated, and the uncertainty associated with the estimation process is saved. The returned atmospheric densities for the CHAMP and GRACE satellites show a marked improvement in the RMS values when compared to the accelerometer derived densities. Next, a method of validating the ANDE results is examined. By examining both the uncertainty in the atmospheric density estimate and the error as compared to the accelerometer derived densities for both the CHAMP and GRACE satellites, a scale factor relating these two variables is studied. This method provides a daily scale factor to adjust the uncertainties in the atmospheric density estimate to determine the root mean square (RMS) error for the ANDE satellites. These RMS values are then separated into several geomagnetic and solar activity bins that allow for a better comparison of the results. From this, the effectiveness of the atmospheric density estimation process is evaluated, and the most effective drag coefficient method is selected. In conclusion, three distinct advancements have been made. First, the drag coefficients have been determined for the ANDE satellites using a larger set of separate methods than have previously been studied, including the Cercignani-Lampis-Lord with a series of separate adsorption models. Second, the POE method is altered to allow for these drag coefficients to be used directly instead of estimated. Finally, by investigating the difference in the uncertainties of the CHAMP and GRACE satellites with their RMS errors, an estimate of RMS errors for a satellite without a base truth model are provided for the first time