Heavy-Tailed Sea Clutter Modeling for Shore-based Radar Detection

Detecting targets embedded in sea clutter poses a clutter-modeling challenge for marine surveillance radar. In this paper, the statistical modeling of sea clutter observed by an X-band high-resolution coastal radar at very low grazing angles (1.05°–1.72°) is investigated. The aim of this paper is to identify the best-fitting statistical distribution to the data with particular attention to the application in the detection scenario. The global goodness-of-fit to the sea clutter distribution and the local fit to only the tail region are both evaluated since the detection probability depends on the whole region of distribution while the detection threshold is mainly determined by the tail region. The results suggest that the Log-logistic distribution is optimal to model the whole region of sea clutter distribution while the recently developed K+Rayleigh distribution, which accounts for thermal noise, fits the tail region best. A general method of calculating the expected probability of detection is also derived to evaluate how the global fit affects the expected probability of detection calculation.