Spatial analysis of telematics-based surrogate safety measures

Introduction: Surrogate Safety Measures (SSMs) such as time-to-collision, harsh braking, and post-encroachment time, are widely proposed in transportation science and are fruitful for road safety evaluations when detailed crash data are unavailable. Hence, this study aims to conduct spatial analysis of harsh braking events to explore their adaptability and informative power in a region with low crash counts, as this approach remains briefly addressed in the literature. Method: The analysis utilizes smartphone driving behavior data and OpenStreetMap road network characteristics of 6,103 road segments in the Region of Eastern Macedonia and Thrace, Greece. A series of advanced statistical and machine learning models were applied. In addition to developing non-spatial models, the identification of spatial autocorrelation led to the development of spatial modeling techniques to account for spatial dependencies. Results: The number of trips per segment, segment length, speeding and mobile phone use are positively correlated with harsh braking. Conversely, motorways exhibited fewer harsh braking events compared to other road types. Furthermore, the number of trips per examined road segment was the most influential predictor, highlighting its importance as a proxy measure of risk exposure. In terms of model performance, the Spatial Zero-Inflated Negative Binomial (SZINB) model outperformed the corresponding non-spatial model. Moreover, the Spatial Random Forest (SRF) model reduced the absolute values of spatial autocorrelation in the residuals and showed a better fit to the observed data compared to the conventional Random Forest model. Conclusions: Geometrical and behavioral parameters can be combined to meaningfully conduct road safety analysis spatially and proactively, as they are highly correlated with harsh braking SSMs, while the SZINB and the SRF model exhibited better model fit than their non-spatial counterparts. Practical Applications: The study results can assist policymakers in developing appropriate countermeasures to reduce harsh braking in targeted spatial units, thereby enhancing overall road safety.