The exploratory analysis of bivariate spatial point patterns using cross-spectra

Whereas frequency and time domain approaches to the analysis of temporal processes are regarded as complementary, the corresponding options for spatial processes, including spatial point processes, are less widely known. This paper concerns the spectral analysis of spatial point processes consisting of two types of events. The spectral functions needed for such an analysis are two-dimensional analogues of those used in the analysis of bivariate time series. We show how the spatial versions of co-, quadrature, amplitude, phase, coherency and gain spectra are defined and interpreted, and demonstrate the forms these functions take under various types of spatial dependence using simulated realizations of patterns with known structure. The potential of spectral analysis to recover information about complex anisotropic structure in environmental point patterns is illustrated using data on the movement of storms in the Sahei region of Africa.