Segment-Based Count Regression Geospatial Modeling of the Effect of Roadside Land Uses on Pedestrian Crash Frequency in Rural Roads

Pedestrians are one of the most vulnerable road users that are prone to injury and death. Various factors have been incorporated into transportation systems in order to improve pedestrian safety in recent studies. The main objective of this study is to develop a segment-based micro-level geospatial-based approach to find the interaction between the frequency of pedestrian crashes and roadside land uses in rural roads. The proposed approach uses geospatial functions for extracting contributing factors and two different lengths of road segments as analysis units to reduce the randomness of the crash locations. These spatial factors are used to estimate the number of pedestrian crashes in each segment using four count-based regression models, including Poisson, negative binomial (NB) regression models, and their zero-inflated extensions. The latest four-year reporting crashes and land use data for a four-lane divided rural multilane in Guilan province, Iran, were tested to illustrate the models' accuracy and performance in the proposed approach. Modeling results highlighted the superiority of the Poisson regression model and its zero-inflated extension for two different strategies of segment length. Moreover, the results showed that residential, commercial, governmental, institutional, utility, and religious land uses have various decisive impacts on the increase of pedestrian crash frequency. This information could be used in long-term transportation systems planning, which would lead to an improvement in pedestrian safety levels.