Angular Profile Estimation on Phased Array Weather Radar by Maximizing Marginal Likelihood

This paper proposes Maximum mARginal Likelihood Estimator (MARLE) for retrieving angular profile of volume targets such as precipitation on phased array weather radar. MARLE aims to achieve super resolution and unbiased estimation simultaneously, which is difficult for existing adaptive beamformers or sparse reconstruction methods. MARLE is based on maximization of a marginal likelihood which mathematically tends to produce sparse solutions to accomplish super resolution. Furthermore, the marginal likelihood is derived based on Bayes' theorem with probabilistic properties of radar observation are considered, where is no technical assumption to produce biases. We tested MARLE with simulated radar received signals and compared it with existing methods. In the simulation, angular profiles of weather measured by CSU-CHILL radar were used as reference. In addition, the other simulation supposing three targets existing closer than mainlobe width of a phased array antenna was implemented. In these simulations, MARLE successfully worked as a super resolution and unbiased estimator.