Towards a safe and sustainable mobility: Spatial-temporal analysis of bicycle crashes in Chile

Highlights

• High local clustering of crashes is observed in rural areas of the Maule region.

• Talca, Curicó, and Linares are communes with high crash clustering intensities.

• Linares has the highest clustering intensity of alcohol-related crashes.

• Large number of spatial co-occurrences between the variables are apparent in Curicó.

Introduction

Bicycle use has proven health, economical, and environmental benefits, in addition to mobility, equity, and socioeconomic access improvements (de Hartog et al., 2010; Rios et al., 2015; Beck et al., 2016). Cycling has commonly been used for physical or recreational purposes, and in some countries, is employed also as a mode of transportation for commuting purposes (Billot-Grasset et al., 2016). Thus, cycling has progressively increased worldwide in recent decades, while having an important impact on the reduction of traffic congestion, air pollution, and carbon emissions, and thus, contributing to a higher quality of life and sustainability. For example, in New York City, bicycle commuting increased by approximately 2.7 times between 2005 and 2017 (Hu, 2017; Klassen et al., 2014). Similarly, Amsterdam has shown a steady increase from 25% in the 1970s to 48% in 2013 (Frame et al., 2017). The city of Perth in Western Australia perceived a growth rate in the number of bicycle commuters of nearly 30% between 2006 and 2011 (Waddington, 2017). The use of bicycle in Latin America has been increasing significantly in recent years, adding up approximately 2500 km of bicycle infrastructure among 56 major cities in 2015 (Rios et al., 2015).

In Chile, over 53% of the population has a bicycle, and there has been a 10% annual increase in the usage of the bicycle for different purposes in the last few years (CONASET 2018). Although the level of bicycle ridership is relatively low (7%) when compared to developed and bicycle-friendly cities such as Amsterdam, Brussels, and Copenhagen (15%–18%), the capital city of Chile, Santiago, is the second city in the region with most daily cycling trips after Bogotá, Colombia (Rios et al., 2015).

Chilean government efforts have focused on improving the bicycle infrastructure, particularly in Santiago. However, Chileans have reported that a poor cycling infrastructure and low cycling lane-sharing exposes them to being involved in traffic crashes, which decreases the likelihood of using a bicycle as their main mode of transportation (MMA, 2018). Chilean normative obligates the use of some safety elements, such as helmet, and frontal and rear lights in urban areas (BCNC, 2018). Therefore, bicycle safety concerns remain a barrier to increased cycling participation. If the society perceives an increase in bicycle safety, then the utilization of bicycles would eventually increase and a sustainable development of transportation would be promoted (Zhou et al., 2019).

Cyclists are vulnerable road users that are exposed to high injury or fatality risks when involved in traffic crashes (Loo and Tsui, 2010; Wegman et al., 2012). Studies in different countries have concluded that the fatality risk when riding a bicycle is higher than driving a vehicle (Raihan et al., 2019; de Hartog et al., 2010). In addition, the number of cyclists killed or injured worldwide has been increasing over the years (Yan-Hong et al., 2010; Billot-Grasset et al., 2016; O'Hern and Oxley, 2018; Useche et al., 2018). In Latin America, the cyclist fatality and injury rate per 100,000 population range between 0.04 and 1.2, and 2.2 and 43.8, respectively (Rios et al., 2015). In Chile, bicycle crashes represent approximately 12% of the total number of annual crashes and are the third type of transportation mode with most collisions after motorized vehicles and pedestrians. As a result of such crashes, on average 179 victims are killed every year nationwide, representing a share of 6% of all fatalities (IRTAD, 2018). Additionally, the proportion of casualties and severe injury outcomes due to bicycle crashes increased from 6.2% in 2008 to 25.2% in 2016. In particular, Santiago has over 40% of the Chilean population and the largest number of daily cycling trips in the country (Rios et al., 2015), causing the largest share of bicycle crashes and KSI nationwide (CONASET, 2018).

Bicycle crashes have been studied in the literature using different methodologies such as Negative Binomial, Bayesian conditional autoregressive logit, Poisson lognormal random effects, and multi-level logistic regression models to determine the relationship between bicycle crash-related attributes and statistically significant causal factors (Kondo et al., 2018; Raihan et al., 2019; Schepers et al., 2011; Siddiqui et al., 2012; Williams et al., 2018). Other studies have employed spatial statistical models to determine the association between different types of bicycle crash variables by taking into account both bicycle crash locations and their surrounding attributes (Hu et al., 2018; Lee et al., 2017; Loo and Tsui, 2010; Morrison et al., 2019; Rebentisch et al., 2019; Vandenbulcke et al., 2014).

Among the existing spatial statistical approaches, spatial autocorrelation has been used in traffic safety data to measure the extent to which the value of an attribute at a given crash location influences the attribute values at neighboring crash locations (Chen et al., 2019; Dezman et al., 2016; Khan et al., 2008; Mitra, 2009; Pour et al., 2018; Xie and Yan, 2013). Spatial autocorrelation assesses the interdependence (or co-variance) of a specific attribute throughout space, which reveals the importance of studying locations with high crash attribute values. Statistically significant clusters of crash attributes are identified, when spatial dependence (similarities) exist among such attributes (Mitra, 2009).

Due to the important features and proven satisfactory results of spatial autocorrelation, the safety problem of bicycle crashes has been widely addressed using this method (Chaney and Kim, 2014; Lakes, 2017; Loidl et al., 2016; Vandenbulcke et al., 2009). For instance, Vandenbulcke et al. (2009) employed Moran's I to detect spatial patterns of bicycle use and casualty risk due to accidents in Belgium. The results of the study identified communes with clusters of safe and unsafe bicycle use, and high and low risk of casualties to cyclists. Chaney and Kim (2014) determined spatial patterns and clusters of bicycle collisions in Cincinnati, Ohio using global and local Moran's I statistics. The authors identified significant high clustering of these crashes in the downtown and southwest areas of the city and concluded that 10 out of 51 neighborhoods presented a significant patterning effect. In another study, Loidl et al. (2016) obtained spatial and temporal patterns of bicycle crash prevalence in Salzburg, Austria for different aggregation units. In the study, the authors concluded that crash locations are highly clustered mainly along bicycle corridors of the city. Lakes (2017) performed a spatial analysis to detect significant hotspots of pedestrian and cyclist crashes that occurred during the 2011–2015 period in Berlin, Germany, and concluded that there are spatial clusters of such crashes tend to strongly concentrate in the inner city area.

Although the study of risky areas with the high numbers of crashes and KSI is very interesting, the inclusion of the spatial component in the analysis of traffic crashes helps to understand the associations of different crash-related factors over space. The interdependence among crash attributes obtained through spatial autocorrelation helps to identify statistically significant aggrupation of the most relevant crash attributes and road safety measures (Mitra, 2009). Thus, this research contributes by identifying which locations with high occurrence of bicycle crashes and their related attributes are recurrently being clustered over time.

The objective of this study is to identify statistically significant spatial clusters of bicycle crashes in Chile that have persevered over time and to determine the existence of spatial dependence among crash attributes and road safety measures. Global and local spatial statistics are used to identify and measure the strength of these spatial clusters. In addition, a spatial co-occurrence analysis is conducted to determine whether recurrent clusters of different bicycle crash attributes coincide spatially with clusters of crash outcomes (e.g., number of crashes and KSI) and crash types such as collisions.

In this study, spatial autocorrelation is employed to explore the clustering of the main contributing causes of bicycle crashes in Chile. In particular, over half of the bicycle crashes and 60% of the fatalities are caused by the imprudence of the driver. In addition, cyclist-vehicle collisions cause nearly 90% of the deaths, predominantly in urban areas. Regarding time of day, the majority of the bicycle crashes occur in the evening between 6 pm and 9 pm at the end of the work shift. A large percentage of the bicycle crashes arise along straight road segments between intersections (CONASET, 2018). These crash statistics suggest the need for studying these and other specific bicycle crash attributes, and identifying the spatial association among such crash attributes over time. Thus, effective countermeasures and public policies are properly targeted toward the improvement of road safety for cyclists in Chile.