Should We Account for Network Distances or Anisotropy in the Spatial Estimation of Missing Traffic Data?

Abstract

In light of the unavailability of traffic volume data for all road segments, the scientific literature proposes estimating this variable using spatial interpolators. However, most of the methods found use the Euclidean distance between the database points as a proximity measure, in addition to ignoring the anisotropy of the phenomenon. Thus, the objective of the present study was to apply Ordinary Kriging (OK) with network distances and anisotropic OK in traffic volume data on highways in the state of São Paulo (Brazil), comparing its results to the traditional isotropic approach with Euclidean distances. Goodness-of-fit measures confirmed the good performance and better suitability of OK with network distances over the analyses that use Euclidean distances. Addressing the anisotropy of the traffic volume data also helped to improve the results. The proposed method can effectively support estimating traffic volume in segments without flow data.