Investigating the spatial heterogeneity of drunk-driving events in Beijing based on a hybrid method with LISA and GeoDetector

Abstract

Drunk-driving events can be divided into drunk-driving crashes (DDCs) and non-crash drunk-driving events (NDDEs). Strong spatial heterogeneity, which consists of spatial local heterogeneity and spatial stratified heterogeneity, exists in drunk-driving events. Therefore, this article proposed a hybrid method with LISA and GeoDetector to investigate the spatial heterogeneity of drunk-driving events. First, an integrated method of Getis-Ord Gi* and Anselin Local Moran’s I, which belong to LISA, was applied for the analysis of spatial local heterogeneity. Then, based on the improved GeoDetector with a highest q value, this article explored the main factors contributing to spatial stratified heterogeneity. Results show that DDCs are concentrated in the urban areas and the sub-centres located in the east of suburban, while NDDEs are predominantly found in the urban areas and the sub-centres located in the northwest and southwest of suburban. The spatial stratified heterogeneity analysis, conducted using the Factor Detector, reveals the pivotal factors influencing the spatial distribution of DDCs. Entertainment venues, alcohol outlets, and population density exhibit significant impacts, with q-values of 0.233, 0.227, and 0.203, respectively. Similarly, for NDDEs, the most influential factor is the density of entertainment venues, with a q-value of 0.243. When exploring the interaction effects between factors, the study emphasises that the interaction between population density and alcohol outlet density shows the highest q-value of 0.787. This highlights a substantial impact on the spatial formation of drunk-driving events. These findings offer valuable insights for implementing targeted measures to improve road safety and reduce the occurrence of drunk-driving events in Beijing.