SPATIAL ANALYSIS OF THE ROAD TRAFFIC ACCIDENT STATISTICS IN THE PROVINCES OF TURKEY

The aim of the study is to describe and model the spatial distribution of the road traffic accidents (RTAs) rate with the factors considering space-time relationship for the period between 2013-2018 in the provinces of Turkey. The RTA rate is modelled with the factors which are population, the number of different types of motor vehicles registered, lengths of three types of provincial roads and these factors measured in the neighbours. Firstly, spatial maps are used to demonstrate the spatial variability of RTAs in Turkey. Global and local spatial autocorrelation analyses are conducted to demonstrate whether the RTA in provinces with high rates of accidents show similar clusters. Spatial regression and panel models are considered a solution to examine the space-time relationship between the RTA and RTA's neighbourhood characteristics. We found that the RTA rate is not distributed randomly across Turkey. Spatial distribution of provinces with high rates of accidents is non-random (Moran's I changes between 0.52 and 0.59 with p <0.001). Moreover, while LISA analysis demonstrates the provinces determined as local clusters, the fixed effects models with different spatial structures show that the RTA rate is positively correlated with number of cars, vans, private vehicles and length of asphalt roads, other factors are negatively correlated and also non-asphalt road is not significant to explain the RTA rate. On the other hand, spatial parameters are significant in all models (p <0.1) and neighbouring region characteristics in terms of explanatory variables do not affect the explanation upon the RTA rate.