Modelling sea clutter using conditional heteroscedastic models

In this paper a class of conditional heteroscedastic models is introduced in the context of sea clutter modelling. In particular, an Auto-regressive (AR) process driven by conditional heteroscedastic (CH) errors (AR-CH model) is proposed as a model for the time evolution dynamics of the modulating component of sea clutter. The CH process parameters of the AR-CH model determine the weight of the tails of the marginal distribution, while the AR component largely determines the correlation structure. Different functional forms of conditional variance models are investigated using real sea clutter data.