Abstract
To detect spatio-temporal pattern of crime clusters/hotspots, the possibilities of three-dimensional mapping methodologies for crime event data are explored using two approaches: three dimensional kernel-density mapping using a volume rendering technique and visualisation of cylindrical significant clusters that can be detected by space-time scan statistics. Both approaches are intended to visualise spatio-temporal domains with high densities of crime in a three-dimensional space composed of two geographical dimensions and one time dimension. The proposed three-dimensional mapping methodologies are evaluated through application to a dataset of snatch-and-run offences in Kyoto City during the period 2003-2004. The results are summarized as follows:
(a) Three-dimensional crime mapping enables effective visualisation of the geographical extents and duration of crime hotspots simultaneously. This method is particularly useful to identify geographical diffusion and movements of crime clusters/hotspots compared to traditional dynamic analyses of crime mapping using cross-sectional maps with arbitrary time intervals.
(b) In practice, the roles of three-dimensional kernel mapping and space-time scan statistics should be complementary. Space-time scan statistics provide clear-cut domains of crime clusters/hotspots that can be used for secondary analyses, such as evaluation of socio-environmental and temporal characteristics focusing on detected domains. However, we should note that the method assumes cylindrical geometrical-constrains of space-time domains. Three-dimensional kernel density mapping provides fuzzy domains with high densities of crime and a useful basis to assess the validity of the assumption of spatial scan statistics and to investigate detailed space-time sequences of crime clusters/hotspots.
(c) Empirical analyses of the snatch-and-run offence dataset in Kyoto City revealed constant clusters/hotspots during the study period in central Kyoto and around Kyoto Station as well as transient clusters/hotspots around several railway stations in the suburbs. Temporal differences of transient clusters show geographical movements of hotspots from the north to the south via the west. We also identified that outbreaks of snatch-and-run offences alternated between a pair of cluster areas. These results suggesting so-called displacement phenomena indicate the need to monitor crime events and effects of crime-preventive actions in a widespread space-time context.
