Re-examining the nature of radar sea clutter

Conventionally, sea clutter has been modelled as a stochastic compound k-distribution. Recently it was suggested that sea clutter was not stochastic but rather deterministic. More precisely, the system could be described as a nonlinear deterministic process, which exhibited chaotic behaviour. This issue is re-examined. It is demonstrated that the two main invariants used in the previous analysis, namely the 'maximum likelihood of the correlation dimension estimate' (DML) and the 'false nearest neighbours' (FNN) are problematic in the analysis of data from real-world systems. The DML and FNN invariants are investigated for stochastic time series. It is shown that both the invariants lead to the false detection of chaos for stochastic time series and thus are unsuitable for determining the nature of an unknown system, such as sea clutter. An alternative method using the correlation integral of Grassberger et al. is used in the paper for a variety of stochastic systems including those associated with sea clutter. Three sea clutter sets are investigated for different polarisation, channels, data length, wind speeds and for pulse-compressed and non-pulse-compressed data. It is shown that in all cases the data were better described as stochastic in nature. No evidence was found of determinism or chaos in the data.