Simultaneous correction of the time and location bias associated with a reported crash by exploiting the spatiotemporal evolution of travel speed

Abstract Accurate occurrence time and location of a reported crash are critical to effective crash analysis. Although there has been a proliferation of studies that attempt to correct the bias associated with a reported crash, most, if not all, of them focus exclusively on correcting the location bias. In this research, we propose to simultaneously correct the time and location bias associated with a reported crash, which is new to the literature. In our approach, we first follow standard procedures to identify the set of candidate links in the vicinity of the reported crash location. We then develop an integer programming model with a set of novel constraints to identify the candidate whose spatiotemporal evolution of travel speed is most congruent with the occurrence of a crash. We subsequently use the time and location where travel speed begins to drop to correct the bias associated with this crash. We prove that the spatiotemporal impact region, which characterizes the evolution of travel speed, estimated by our model is consistent with the propagation of shockwaves even when there are multiple candidate links and the exact occurrence time and location of the crash are unknown. This relaxes the standard assumptions required by existing models in the literature. We validate our model using real crash data in Beijing and find that our model can reduce the average bias in time from 7.3 min to 1.6 min, or a 78.08% reduction; and reduce the average bias in location from 0.156 km to 0.024 km, or a 84.62% reduction.