Impact of Data Selection on the Accuracy of Atmospheric Refractivity Inversions Performed over Marine Surfaces

Within the Earth’s atmosphere there is a planetary boundary layer that extends from the surface to roughly 1 km above the surface. Within this planetary boundary layer exists the marine atmospheric boundary layer, which is a complex turbulent surface layer that extends from the sea surface to roughly 100 m in altitude. The turbulent nature of this layer combined with the interactions across the air-sea interface cause ever changing environmental conditions within it, including atmospheric properties that affect the index of refraction, or atmospheric refractivity. Variations in atmospheric refractivity lead to many types of anomalous propagation phenomena of electromagnetic (EM) signals; thus, improving performance of these EM systems requires in-situ knowledge of the refractivity. Efforts to inversely obtain refractivity from radar power returns have done so using both reflected sea clutter and bi-static radar approaches. These types of inversion methods are driven by radar measurements. This study applies a bi-static radar data inversion process to estimate atmospheric refractivity parameters in evaporative ducting conditions and examines the impacts of radar propagation loss data quantity and source location on the accuracy of refractivity inversions. Genetic algorithms and the Variable Terrain Radio Parabolic Equation radar propagation model are used to perform the inversions for three refractivity parameters. Numerical experiments are performed to test various randomly distributed amounts of synthetic data from a 100 m altitude by 60 km range domain. To compare the impact of location of data on the inverse solutions, three domains were examined from which data was sourced, including the whole domain (0 m to 100 m altitude and 0 km to 60 km range), a lower domain (0 m to 60 m altitude and 0 km to 60 km range), and a long-range domain (0 m to 100 m altitude and 30 km to 60 km range). Comparisons of inversion performance across experiments involved evaluation of several metrics: fitness scores, fitness-distance-correlations, the root-mean-square-errors of refractivity profiles, and percent errors of each individual refractivity parameter. The results of the data quantity experiments show that propagation loss measurement coverage of approximately 1% of the prediction domain yields the most accurate refractivity estimates. It is concluded that this amount of data is needed to sufficiently eliminate non-unique solutions that were observed using smaller data quantities. The results of the regional study indicate that the long-range domain produced slightly more accurate results with less data compared to the other regions. From the results of these experiments and prior studies, four specific sampling patterns were developed that were hypothesized to generate accurate inversion results. It was shown that the pattern containing the most data cells with the widest spread over the domain generated inversion results with the highest parameter and refractivity accuracy; although, a second pattern that sourced data concentrated in a short range low altitude region performed similarly with significantly less data. The results from this study enable advancement of refractivity inversion techniques by providing insight into where and how many EM measurements are needed for successful refractivity inversions. Improvements in refractivity inversion techniques enable performance improvements of EM sensing and communication technologies