Road traffic crashes remain a persistent global challenge, causing millions of fatalities each year and severely injuring many more. Beyond their human toll, crashes impose significant socio-economic costs worldwide, yet traditional approaches to identifying hazardous locations often rely on simple frequency-based metrics, which fail to capture spatial dependence and variations in crash severity. However, crash events are not spatially random in many urban contexts; instead, they exhibit strong clustering patterns driven by recurring geometric, operational, and behavioural factors.
Identifying the spatial distribution of traffic crash hotspots provides valuable insights into the root causes of crash occurrence over the area under study. This knowledge supports decision-makers in assessing road safety risks and implementing targeted countermeasures to reduce crash rates. Nevertheless, identifying areas with high crash potential remains a complex task.
In this regard and within this study, we aim to address these limitations by developing a comprehensive spatial analytics framework for assessing crash severity patterns. The proposed methodology integrates severity-weighted crash indices and applies a suite of spatial statistical techniques to detect clustering behaviour and identify micro-scale hotspots. Specifically, the framework employs Average Nearest Neighbor Distance (ANND) to evaluate the statistical significance of spatial clustering, Global Moran’s I to assess overall spatial autocorrelation, and Local Moran’s I to identify significant local clusters, including High-High (HH) and Low-Low (LL) severity areas, as well as spatial outliers.
The framework is designed and developed within the QGIS environment as an experimental plugin, offering a user-friendly interface and a streamlined workflow for performing advanced spatial statistical analyses. The solution is scalable and transferable to different geographic contexts, provided that appropriate input data are available. The methodology is demonstrated through a case study in the greater Limassol area in southern Cyprus, where it is used to analyse crash patterns and evaluate spatial risk.
Results reveal strong clustering of crash locations based on ANND analysis, while Global Moran’s I indicates weak or non-significant global spatial autocorrelation of crash severity. This suggests that severe crashes do not form broad regional patterns but instead emerge as highly localized phenomena. Local Moran’s I analysis confirms the presence of small yet significant High-High severity hotspots and several spatial outliers (High-Low and Low-High), typically located along major arterial corridors, network transition points, and mixed-use zones. A comparative analysis of pre- and post-COVID periods shows a substantial reduction in crash frequency and cluster intensity, indicating measurable impacts from altered mobility patterns and the introduction of automated speed enforcement measures. Overall, the findings highlight the importance of localized, severity-aware spatial analysis for modern road safety management. The proposed framework provides actionable insights for black-spot identification, targeted interventions, dangerous goods routing, and integration into digital twin and intelligent transport systems (ITS) decision-support environments, supporting data-driven strategies aligned with Vision Zero objectives.