Abstract

The homogeneity of road networks, in terms of network structure, can be used to separate road networks into neighbourhoods with distinct road network patterns. Neighbourhoods that differ in terms of road network structure may also differ in terms of accessibility to opportunities such as jobs and education, or facilities such as clinics. This is of crucial importance especially in the developing world, where informal road networks can arise when residents create their own roads without the government being informed. Thus, detecting road network homogeneity provides valuable information for sustainable development. Homogeneous road networks typically contain roads in one of two directions, namely the major direction and its orthogonal direction. Thus, modelling the direction in which roads occur in a network, from some reference point, could provide valuable information to assess road network homogeneity. In spatial statistics, road networks can be represented as spatial linear networks. However, there are currently no methods within spatial statistics to test for significant directions of lines within spatial linear networks, nor to cluster roads by their direction, nor to assess any such clustering in terms of statistical significance. This research proposes initial methodologies to detect the main directions present within a road network, cluster roads by their directions, and test for the statistical significance and goodness-of-fit of these clusters. This is done by integrating directional statistics with spatial statistics in a novel manner.