Assessment of the Methodology for Bounding Loran Temporal ASF for Aviation

As we become increasingly dependent on GPS for many position, navigation, and time (PNT) applications, it becomes increasingly important that we have an alternate means of obtaining those capabilities. This is especially important in critical applications such as aviation. As part of the ongoing Federal Aviation Administration (FAA) Loran evaluation, the system is being assessed for its ability to provide stand alone navigation capability through all phases of flight. This will allow users to retain much of the functionality enjoyed when using GPS should the system become unavailable. For Loran to be used for aviation navigation, it must meet strict integrity requirements. A key requisite for meeting integrity is the ability to bound position errors to a high degree of confidence. This, in turn, means bounding range domain errors and variations. One major source of variation is the temporal variation of propagation delay known as additional secondary factor or ASF. Models were developed to provide the bound on the ASF variations for the assessment of Loran aviation coverage. These bounds are designed to meet the integrity requirements on the error. Significant amounts of high quality Loran data from the U. S. Coast Guard (USCG) has been used to generate the model. This, however, left little data of adequate quality to perform the independent assessment of the models’ performance. Hence, little or no independent validation of the efficacy of the bounds could be done until recently. This paper presents an analysis of the models for bounding the temporal variation of ASF. It uses data collected in 2006 and 2007 from seasonal monitors installed by the FAA Loran evaluation team. It will examine both the performance of the bounds in the range and position domain. One major range domain bound issue examined is ensuring the ASF bound attributed to the uncorrelated component is adequate. This is important as the uncorrelated component is treated much more conservatively than the correlated component. Additionally, some sensitivity will be examined to test robustness of the model and implementation.