Improved Coordinate Registration for Over-the-Horizon Radar: Sampling-Based Approaches

In the coordinate registration of skywave over-the-horizon radar (OTHR), errors arising from ionosphere modeling and parameter estimation are the main sources. We propose reducing these errors by leveraging reference sources together with the corresponding OTHR measurements, and exploiting high-accuracy ionospheric models and computational techniques, including the international reference ionosphere model for the ionospheric parameters, the Gaussian Markov random field model for the spatial variation and correlation of ionospheric activity parameters, and 3-D ray tracing. Formulating the estimation of ionospheric activity parameters as a Bayesian inverse problem, we tackle it within a Gibbs sampling-based framework. First, we propose a coordinate registration method that incorporates importance sampling and resampling, utilizing an exact ray tracing model. Moreover, acknowledging the significant computational overhead caused by the exact ray tracing, we propose a local approximate sampling method for the complex forward model of ray tracing based on multimodel learning in the Gibbs sampling. Here, the quasi-Monte-Carlo method is used to draw a few samples from the exact forward ray tracing model, which are then used to learn the approximate model. We use samples from both the exact forward model and the approximate model to estimate the posterior probability distribution of ionospheric activity parameters. Simulations indicate that in the case with densely distributed reference sources, the accuracy of OTHR's coordinate registration using our proposed methods improves by 88.47% compared to the a priori estimates. Even in the case with sparse reference sources, the accuracy improvement remains substantial at 68.01%.